Day 2 - International Conference on Trustworthy Al for Intelligent Computing System ICTAICS 2026

本站添加: 2026-06-01

[00:04:35]Bloom and u uh even somebody is from nontechnical background they will be also using it for their day-to-day life even it is academic writing or it is just a email or even you want to do a WhatsApp also that time also what we do we go to charg we go to the cloud we go to the gemini we type we make the images and then we get the things done there.

[00:05:00]So you see we we are not limited to the agentic or I'll say we are not limited to the LLMs now we have more than that and in today's this talk we will we will figure it out that what is the future what we are talking about uh can we can we just relying on to the LLMs now the answer is no but what is the replacement and what we are doing uh for today and you see if you want to become a job uh job ready or industry ready so You need to be up get updated with the latest technologies and even uh we we talking about it today we are not ensuring anything in next month we have something a new model appearing up and that is doing wonders in the market so we need to keep upgraded we are from the technology background we are from the science background so this is something which is required at our perspective uh so when when I do say uh the same concept here so you see uh let just give me my brief uh introduction why I am onto this topic. I am working with CGC University as associate professor and u my one of the student she approached me for this agentic AI and uh that is the point you know I started figuring out and today um I I want to do everything into the same domain and I want to explore it more. So this maybe this uh session give you motivation to go ahead on to agentic rag or agentic AI. So when we say when we talk about this evolution here or when we do talk about uh how how it is started so you see we we used to go through the machine learning then deep learning is there advanced deep learning models were there pre-trained models u appeared over there. So when the pre-trained models were there at that time what we what we were having okay we we have something uh like before that it was a rulebased system static knowledge limited adaptability then LLMs are there generative AI content generation natural language understanding you give a document you get the summary you ask a question you get the answer uh you want to translate a language that is also being done so after that we have something Now agentic AI so might be most of you are not u you know are not good to go with the agentic AI but you see believe me we have more than that so uh get ready buckle up your seats and uh get to know about the these things so agentic AI is more related to a goal oriented so just not directly asking question to the LM and getting the answer it's more than that uh it's a goal oriented approach and uh we we can have the decision making we can have this in the autonomous section also. So we can apply automation. So we we will see how it is happening. Then the rag is there. So rag is what in a simple language if I do talk about it is basically a retrieval process. But in that retrieval process you do see um I am providing a knowledge base also. For example I do say here that this is my this is my company's HR policy. This document I am giving it to you. You use your knowledge, you use this documents knowledge and then give me the answer. So you see in a simple language what I am doing I am connecting my LLM to a knowledge base within the form of a documentation within a form of some guidelines or any any uh apart from that. So once I do that that is what that is my retrieval process and we do call it rag here. So uh why why we are coming to this why we are talking about this here. So you see uh when when I do talk about here so in this case uh when I say that there should be some problem that is why we are coming to the solutions there. So you might have heard about the hallucinations. So what is hallucinations? So LLMs are uh seriously affected by the heloc hallucinations and the reason being uh you see you give a scenario and it starts getting the answer it is trying to fabricate the answers there and uh whenever you have these LLMs there you can't expect you can't expect that uh you know this knowledge will be updated or something so for example today you are searching and there is some news that appeared yesterday it's not shortity that you'll get that trained data because it is training on to a global manner not onto the local manner. So that is why the fabricated information the knowledge uh limited knowledge I could have. So we could not have any access to the private uh data. We have the limited reasoning across multiple sources. For example, I do ask what are the latest cyber security threats in 2026. So this will be giving me wherever it has been trained off it. it is not giving me the latest answer and I might think it it is a latest answer but this is this is something I'm getting a outdated knowledge here uh so let's come to the rag now so rag stands for retrieval argumented generation retrieval agumented generation we are doing so you give a query it retrieves the document relevant documents so for example I'm asking about that what is the leave policy maybe this leave policy is already described in the HR policy so I am agent we are calling it agent so agent is responsible I am not providing anything this agent will be finding that document and it will get the answer accordingly and finally I'll have the answer so you see now what we are using we are using llm also we are using these additional source of information also so what we get get in the last uh accuracy is improved we can get a do domain specific knowledge might be uh my charg is not trained for the Indian cuisines or any any another thing. So maybe I can provide some document and I can get the answers directly onto this. So we can have the reduced uh hallucination also up toate information we can uh we can talk about. So when I say a basic architecture when we say so through the LLM we have the query we have the retrieve knowledge and that we are getting it with the ra architecture and we are getting the answers onto this. So we we call this document is knowledge base. Uh we could have some model to apply onto that because you see whenever we have the problem we have some solution with us. So if agentic rag I'm talking about so it means this traditional rag is suffering with some problem. So you see this is just a single retrieval step might be I have multiple things to do. I don't have any planning capability there. uh it can't take any decision and it is it is not able to interact with multi multiple tools. Now when one example I'll give uh we have different different APIs to fetch the data from the web. We call it web scrapping or something like that. So this rag is not able to do that. So you see these challenges are actually the key points of agentic rag uh that we are seeing here. So in this case if you uh if you do see so I could have something on to you know no planning capability for example I say uh analyze vulnerability reports and create a mitigation tickets. So you see this traditional rack would not able to do it will not it will give you just a fabricated answer. You will not get anything uh on top priority of this. So this is you see this is very very important. Now this leads to emergence of agentic AI. So now you see now let's try to understand what is agentic AI. So agentic means one agent is working for everything for planning for calling from execution. So you see this is a automation uh system which is perceiving identifying the reason also after identifying the reason it is planning.

[00:13:07]What it is planning? So you see um we do we do see here that I I booked the event manager. So this event manager person is responsible for now 0 to 100. So it is going to uh tell me about the goods also raw materials implementation venue booking everything is taken care by the event manager. So similarly now I have a agent AI which is responsible to call any another tool to plan to reason to uh you know to implement. So everything is now being taken care by the agent care.

[00:13:41]Now you see uh when I'm talking about this here u we do talk about the recession is happening uh multiple companies are laying off. So you see how how this is actually shaping the technology how this shaping uh this when I say that how everything is being implemented the planning is happening so you see that manager's task is also being being with agility so the question arises so what what about us so the the simple answer is that we need to update ourself we need to groom ourself we need to be you know familiar with the latest technology and uh how it is happening how everything is working over So this this we should we should learn about.

[00:14:22]Okay. So you see one of the things here uh this AI will be answering the question also and completing the objectives. I'll tell you with the example also. So you see when I say this agentic rag so this is combining now multiple things multiple things you see uh rag which rag we were talking it means that retrieval document process this is also happening planning is also happening. It can also integrate with multiple tools. For example, I need a API for the Google maps, it will be done. I need a API for the weather uh to know the weather that will be done.

[00:14:55]Calculator. So these all things can be done with the agent integr. So you see this manager is so smart. It will be linking up with LLMs also. It will be linking with u with I'll say with the uh uh with the uh rag also with the multiple tools also, APIs also plus it is planning also. And you see you give a task this agentic rag will decide which tool should be called which uh document should be retrieved which API should be connected. So you see this decision also has been empowered with the agent thing.

[00:15:32]So here you see we could have multi-step reasoning also for example you see uh we can u I I say that plan my day. Okay I am going out onto this location. I have a meeting in uh I have a meeting in Delhi. Plan my day. So you see might be requires you know uh booking of the tickets. It might require my hotel booking. It might require my scheduling.

[00:15:54]It might require my PPT making up. So you see these all things can be done easily by the agent integrant. So this is something which we are working now in the many of the industries that we we do see might be you see this this is a turning point as in you might be still struggling with learning the machine learning models that is also required but you please be clear in your mind more than that has already happened and you need to see executive is responsible for weather API one executive is responsible for this. So you see this planner agent will be having multiple other agents. So you see after knowledge retrieval maybe it has a reasoning agent it has something as a tool innovation which tool need to be called and then it will be in the action mode and ultimately we can get the feedback maybe I require some you know uh inputs also now one more thing I don't know you are aware about of this or not uh so usually at least at our times we were having one subject software engineering and uh we used to learn you know project development life cycle we used to and you know how this testing is happening everything but you see uh you can make softwares also with the with the agentic AI or agentic rag and uh we do have a testing module also so you see I given a python code I have asked I have given a prompt through the user query that make a code for this Python code now I can design this agentic rag which can also implement basically your uh the testing module also also so you see that that can be also done. So we don't we we can link up entire product development life cycle with the agent integr. So uh and uh industry is doing this. So the core components you see. So now I'll talk about u what what how we can you know implement these things. So uh I'll be giving you the names you can explore it more. Uh lang graph is one one of the framework. Crewai is one of the framework. Autogen is frame one of the framework. So you can use any one of three each framework is having its individuality and u so you if you see here what is what is this framework so this framework will make you enable to create the agentic graph so in the python we have all the libraries and we can see this and we can implement that and then when I talk about I'm providing some sub supporting documents that can be done in the vector databases um it can be done through the hybrid search then we can connect it with the LLM which LLM you want to connect OpenAI ch GPT you want to connect it to Gemini you want to connect it to the any claw there so this is your call how you want to go so this decision also you can take might be you needed some tools also and then you need some memory also memory because you see you're keeping everything there and you see before you get the answer there is lot of lot of things are happening at the back end now in this case you see when these things are happening onto the back end what I'll do or what I'll go for we need a memory also when I say memory so I could have you know short-term memory or long-term memory so maybe I can take help of cloud also to uh keep everything secure there or if I want to I want to contain a local memory so that also I can do over there so these all things are possible into into these cases okay similarly similarly when when I do See here um planning and the reasoning. So this is your one of the component where you do decide okay which policies which documents you get which vulnerabilities need to analyze. So this you see if you have a task like perform a cyber security risk assessment so that can be done into into the planning format. We could we could have multiple tools which we can use in the agent integr uh APIs are databases are there calculatorsing system email system. Uh so these all things you can do. So maybe you you design something uh that you fetch the data also and after fetching the data email part has been also done. So you see the this this automation has taken a very good step in uh in the further usage. So we were talking about the memory. So memory you can decide like your rag will be using what kind of a memory short-term or long-term short-term means just like charge GPT you can understand your current context is shortterm long-term is your entire history. So anyone of you if you have used the paid version of the charge GPT so that time you might have uh seen that uh peripheralization.

[00:20:42]So you see I am relating it with the LLM so that you can understand but we we are talking about the agentic rag here. So if we compare this aentic rag and the traditional. So traditional is just a singlestep process just a smaller one module. Agentic is multi-step. Uh we can do the planning in the agentic rag only.

[00:21:01]We can use multiple tools in the agentic rag. The combinations are any number of things there. Autonomy is uh high because you know I can use multiple type of data sets also or data databases for example one data is in JSON format another data is in another format so that conversion can be easily done uh okay so if we take a example now so I'll just just take one example here so what is cyber security use case if you see so there is a operation of sock security operation center we call it so we could have something you know uh agent agent will be taking action. Then we we can have uh threat intelligence retrieval, process retrieving. We we can have one agent who is responsible to analyzing the log cases in case some alert is there that can be done. We can have some uh recommendations also. We can have some incident tickets also and this will be giving us the faster response in the healthcare. If you see um I can have one agent which is retrieving medical information then it is seeing the consultation guidelines. It might review some research papers also and then it can generate recommendations also. So this is also you see u the healthcare diagnosis in the enterprises if we say uh for example a employee is giving a query prepare a compliance report for the last quarter. So agent will be again retrieving the policy, collecting some audit data, summarize the findings, uh it is generating the report and it is also emailing the stakeholders also. So you see this can be done um over there and um if you're interested to you know implement the these things you can just explore these things. So I'll I'll be sharing this PPT with the organizers. So we we have in the technology stack when when I do talk about the uh in in this case you see I can have something you know LM open AI GPD we have different models five you want to use you want to go for there uh entropics claude is there Google's Gemini is there to save the databases we have pine cone uh vivid is there kuma is there and these frameworks you see crewi l graph and autogen so you You can decide anything. For example, in QUI framework, I'm using Chroma vector database for implementation of OpenAI GPT model. So you see your multiple agents will be implemented by the CUI for storing that information you will be using Chroma and your LLM which you are using to generate the responses that is GPT models. So that we can see uh but some of the challenges as of now which which we are having there. So maybe that you can explore. So as of now this is I'll say very niche stage and uh when I say we do not have that much speed. So speed when I say you see back end in the back end multiple processes happening obviously it will take more time. So uh that is one of the thing retrieval quality we still need to ensure that uh you need to give that kind of a documents only which which is actually resulting due to the good output over there. Uh similarly you see we could have security risks also because you see data is everywhere and u we need to go through the guidelines might be we still face the hallucination is there and we still needed the human oversight there access control audit trail so everything need to be governed. So we we always say that this artificial intelligence has to be combined with the human intelligence but for that human should be already knowing that what artificial intelligence is doing. So this is this is how we can become smart u in the future if we see so we can have more multiple agent collaboration we can have entire workflow onto the autonomous self-improving system is also there. So you see this rack was just retrieving the information. Agentic rack is more than that. Understanding, planning, reasoning, acting and giving us onto the information uh goal. So uh this is this is it from my side. The one thing I just want to add up here. So you see this aentic rag this we are talking about and uh in in this case it just started with with the uh with the same motivation how we are doing the things. So those who are into the deep learning they already know that deep learning is also being as a motivation of human brain and similarly for the task implementation also we have what we have the agentic rack. So technology is simple just we need to be you know we need to be upgraded to to this and we need to see that how we can take care in this uh in our career in our opportunity in our research. Okay. Um that's it from my side. Thank you. I'm open to the questions if anyone have and meanwhile um I I'll just suggest one thing anyone of you even you are not from the technical background CAC or something. Uh try to use one of the tool lovable. So you'll see how this uh you know how this you know artificial intelligence has taken place. You you can have everything entire website full stack entire full stack development just by your prompt only. Just try it just half an hour process and you you see how amazing tools are there. One of the tool you can see napkin.ai that also you can explore and uh superbase is there. So these tools uh you if you want to actually see that how in the technology it is happening you will uh you will be amazed by the benefits over these Audience, do you have any questions?

[00:27:49]Thank you. No questions.

[00:28:04]on behalf of Amrita Institute of Engineering and Management Sciences Bengaluru and the organizing committee of the international conference on trustworthy AI for intelligent computing systems. I sincerely thank uh thank you Dr. Aishu Sharma our esteemed keynote speaker for your insightful and engaging session. Madam your presentation on AI system evaluation agentic AI the limitations of standalone LLMS and the rag including its architecture and challenges provided valuable knowledge and practical perspectives to all participants. We are grateful for your time expertise and contribution to this conference madam. Thank you once again madam.

[00:28:48]>> Thank you. Thank you so much. Thank you for inviting me. Great to connect. Thank you.

[00:29:00]Uh I can leave, right?

[00:31:43]Uh uh I am sharing my slides. Uh just wait here window.

[00:32:15]Uh now can you see my slides?

[00:32:21]It's okay.

[00:32:25]>> Anybody?

[00:32:26]>> Yes, it's visible.

[00:32:28]>> Okay. Okay. So, thank you and good morning everybody and I thanks organizer for giving me the opportunity to discuss the importance of the mathematical modeling of physological systems and in this particular work. So, you know the everything is related to the system.

[00:32:47]Then we have the signal and we create hypothetical models and then we validate our results. And in validation our result we need the mathematical models in just like a block diagram. And you must have studied in the control system or in some other applications where we have a block diagram or signal flow graph just to give the enhancement of that particular system into a model and then we go for its analysis part and for analysis part we can use LLAS transformation discontinuous system and discrete system and that system can be converted into a discrete system also with the help of some filters and some techniques where we have to convert S the analog signal to J domain also. So, so we can validate our result and in this way we can say your system is good or system is bad. So, uh for the time being uh the system any anything related to system let's screen now.

[00:33:48]So, it is now system. System is the system consisting of various elements.

[00:33:54]They are interconnected each other. They may be independent but whenever they are going to perform a task they will unite together and to give some output and the output may be maybe the set point and in control system we say the set point but in a in a uh physiological control system we say it's a homeostasis just like in homoasis what should be the blood pressure 120 by 80 the heart rate 60 by 80 though it's a it's a homeostasis must be maintained mean set point the system consisting of components and these components are have a some purpose to perform one task and just like in human body and you know the various organ system they are independent but they will work together to perform one task. Task means to maintain the homoasid or the set point.

[00:34:42]So control system interconnection of component forming a system configuration that will provide the desired response and desired response in our control system may be the set point and in a human body it is known as homeostasis and the process under system just like a cardiovascular system, respiratory system, circular system, muscular system, nervous systems all are system that means they are all are process and in this way we can manipulate the So like this particular system. So you must have noticed this in your control system. You must have noticed in your control system. What is this system here? The system means the man-made system is being controlled by the human body. So both are both have to work together. So in this what we are doing.

[00:35:32]So a operator will check the water level. If water level is up to the mark or it has achieved the uh the the limit so through hands through it eyes it will check I mean it's a sensor. So it is a sensor will give the signal to the brain and brain will give the signal to the hand. Hand means the nervous system and muscular system. So you can say the then the operator will close the wall and in this way it can be controlled or it is just a a human body and a and a system.

[00:36:05]So advancement here instead of watching the water now it will check the meter if meter has achieved so it is the improvement. So human operator. So human operator. This is a human operator. Just wait.

[00:36:33]So this is a human operator where it it is acting as a sensor. This is a sensor and this sensor will check the meter level and this signal will transfer to the brain where we have a controller.

[00:36:44]The controller will check and according to that it will give the signal to the muscle. So motor neuron. So ephren signal ephr means outside response outside signal reaching to the brain and the brain will sense whether we have to close or open or stay calm and then it will give to the muscle. So it's a complete feedback system. It's a feedback system. So you can say the man-made system is being controlled by the human body. So simple when you're driving a car. When you're driving a car, this will give you driving a car. So you are driving a car again you have to sense through your sensor is the outside sensor is the eyes. The signal will give to the brain and the brain will give the signal to the pedal or your hand just for the movement. This you know very well and you have must have studied in control systems.

[00:37:36]So simple one one thing related to let's identify the organ system components and the input and output and of the biological control system consisting of a human being reaching to an object.

[00:37:50]Suppose you want to take something and you want to just hold the class. So how you will go that this is same the sensor will give the brain and brain will transfer the signal to the your the hand. So in this way you can the ephrine neuron ephrine neuron outside sensor inside sensor it will be like this. So what are the major organ system in a human body? Major organ system we have a skin system. The largest organ of our human body a very good sensor non-invasive method. Many thing we can transfer through the brain through through particular skin. Skin is the largest organ system of our human body. Then we have a nervous system, cardiovascular system, respiratory system, skeleton system, endocrine system, digestive uranium and reproductive. So why we are card system?

[00:38:39]Because let it be think about your brain or you can say respiratory system.

[00:38:44]So it consist of various components.

[00:38:47]Cardiovascular system consist of it's a organ is heart. But when it is a consisting of various components it just like a system system will work for a betterment of a human body that is known as the to maintain the homeostasis.

[00:39:03]So just like introduction to human body what we have to study that minimal living organism is the cell structure which we cannot see through neck dice but it can be modeled mathematically we can model and we can simulate because you know whenever we are going for some biomedical application we have to take the ethical clearance for the purpose of validating the result. If you can conver mathematical model, you can simulate your model then you can validate through with the help of a phantom studies.

[00:39:38]Phantom is just like a human body component. So nervous system homoas negative feedback system positive feedback system homoas imbalance means whenever set point is disturbed whenever your homoasis is disturbed 120 by 80 is disturbed that is homeostasis imbalance.

[00:39:54]Survival needs the body communication means the communication between the various components just like a neuron the chemical energy electrical energy and then again chemical energy and transfer of the signal that is communication transport mechanism it may be passive transport it may be active transport internal communication that is inside the human body your inside the human body parameter must be within the limit just like your temperature 37 7 to 41° centiggrade it must be maintained inside whatever is outside maybe 50° maybe 0° - 50° inside must be maintained it is maintained with the help of a your nervous system muscular system parasympathetic sympathetic everything.

[00:40:41]So there are so many application where we can go for mathematical model just like a cardiovascular system. A cardiovascular system consisting of blood vessels control of blood vessel diameter means the arteries blood vessel we can it can be controlled internal respiration the oxygen required for internal or organic organ system heart blood flow through the heart blood supply to the heart. So it is a a a cardiovascular system for a human body not a fetus because I will discuss the fetus it is a for a when the heart is working your respiratory system is working and you it's a fully developed human body where they will work so your heart is working everything is working because the heart in the fetus and some other parameters some organ system are not working till the birth. So that is why we have a very important mathematical model for pto maternal monitoring.

[00:41:39]You can see the this particular heart it's a pumping action. So this pumping means it's a you can see this is a heart organ is hard but it is known as system.

[00:41:50]System is this is a one in a lay man. In a lay man you can say it is a pipe. It's a pipe or maybe a inferior minava and inferior minava. As a layman it is a carry it carrying the deoxxated blood.

[00:42:06]The deoxxy blood from the upper side of the heart and the lower side of the heart will enter to the chamber number one. Whenever we are going for mathematical molding the chamber may be storage element may be considered as a capacitor. The capacitor means charging element and when this blood is going toward this particular second chamber.

[00:42:23]This is a through wall and this wall in electrical may be a diode one way transfer. This blood will go to the through this particular wall and going towards the respiratory system. The respiratory system is sec second or second organ system or respiratory system where there are so many components and where this def blood will be converted into oxy blood and this point at this point the blood is entering to the third chamber with the help of a this blood vessel through respiratory system and through respiratory system this blood will go to the last chamber through this particular wall and this wall will give you blood will exert the blood from this particular wall to this aerota and so you can say that there is a one pump because it has to be transfer the blood the oxygen blood against the gravity this also against the thickest part of the heart is your this is left ventricle so in this way you can model if you want to make a model so it can be converted mathematical model then we can simulate this particular system so it is a independent organ system your respiratory system is independent organ system. But whenever they are they are going for a one noble cause, noble cause means to maintain the blood supply to give the oxygen to the each and every element of the or cell of the human body internal system. So they cannot in they cannot work in isolation. They will have to work in in a united form and then they will go for your the it will give the homoasis balance system. It's a closed loop system type. It it is not a openloop system. So you can say this is a circulatory system. This is systemic circulation. Internal circulation. This is known as external circulation.

[00:44:11]External circulation means we have to inhale. Means it is a there is a respiratory system. We are inhaling.

[00:44:19]This inhaling will be converted into oxygen blood. So it is internal respiration system. It is external respiration system because this will give you the explain explanation of the pto maternal monitoring. This a develop arter heart.

[00:44:35]So in this way this is a the circulation the complete flow of the blood and this blood for a human body not a fetus it is for a human body.

[00:44:47]So in this way if you want to go for mathematical model so you can say it is mathematical model everything has been converted into a block diagram and each block diagram let's say chamber number one array you can consider is a capacitor when the blood is flowing from chamber number one to chamber number two against the wall you can say the resistance opposition of the blood flow blood may be thin blood may be thick depending upon the position so in this way you can create a model so this is a complete block diagram duction So where we study this in our in control system.

[00:45:19]So you can simulate this.

[00:45:22]So with the help of a log diagram duction or a signal flow graph the analog system that is HS the transfer function of this particular system can be given into the discrete system and we have a discrete system advantage over analog system. So it is a complete block diagram for the cardiovascular system.

[00:45:47]See this is the complete program of a human body consisting of consisting of your cardiovascular system and consisting of respiratory system. Here we have a nervous system also the nervous system will control the blood pressure of the system. Whenever there's a increase or decrease in the blood pressure this blood pressure will be checked by the your aota where we have a bar receptor sensor and camera receptor sensor will give you the output based on the parameter. This is your system. So here the nervous system this because nervous system is very very important for understanding the AI that I will give one slide post this also nervous system you can just go through this neuron properties of the neuron cell body axon the nerve impulse type of nerves snaps and neurotransmitter central nervous system fluid and pericular nervous system this everything will be used in your EI system. So I will give you the how it is being converted into AI system.

[00:46:49]This is your AI. This is your the brain nervous system. The central nervous system. This is particularly the sensor.

[00:46:57]This is ephrine neuron. Sensors maybe sight, hearing, smell, taste, internal camera receptor, gamce receptor, better receptor will give the signal to the brain and brain will sense and will give the motor neuron different signal will affect the skeleton muscle. Skeleton muscle will give you the strength.

[00:47:14]Skeleton machine will give you the give you the power. The skeleton machine give you the maybe give you the shivering for controlling the temperature of the human body. If there is a temperature is less it will give you the parasympathetic signal will increase the blood pressure. Heat will be stored and in this way you can control the your human body will control the temperature. So this is and this is respiratory system consisting your nose and n position structure the various organ system in the respiratory system.

[00:47:43]So that way it is known as otherwise system is or the organ is the lungs and it is known as respiratory system because of consisting of various components. This is respiratory the nasal lary fics like this this consisting of system because it has a various part. It may be your controlling the vocal cord. It will will control the breath rate everything it will give you oxygen house.

[00:48:08]So difference between the physiological and engineering system and engineering control system is designed to accomplish a given task with set point.

[00:48:16]Physiological systems are versatile and capable of serving different function or to maintain the homeostasis. Here we study the in physiological control system we study the various organ system its component and working physiological system means they are physiology means giving the signal. A system will give you the signal and that signal we will store and we will validate the our system behavior with the help of a signal.

[00:48:46]So this is your this is already I explained the cardiovascular system and cardiovascular system how this cardiovascular system will work already I explained.

[00:48:56]So definition characteristics and function the cardiovascular system it is a fluid closed loop system connecting all system will give you the blood it's a closed loop system it cannot be open system everywhere it is through a passive transport system this will give you the oxygen level it will reduce the it will means carbon dioxide will be out like this so when we inhale when we inhale carbon dioxide is 78% sorry uh When we inhale this is oxygen is 21%.

[00:49:30]And carbon dioxide is 04 and when we exhale is 4% 100 times the car carbon dioxide we inhale. So you can say that how much carbon is inside our human body and how much it is consumption and your cell will give you output carbon dioxide. That is very very important parameter for monitoring or for sensing the carbon dioxide whenever we exist.

[00:49:50]That is very very important.

[00:49:52]It is not a mathematic. See this is a mathematical mode.

[00:49:56]So this is mathematical model. This is a pipe one chamber. It is a capacitor resistor capacitor like this. So it is for this particular. So when we are converting this to transfer function see this RA this is the input output with the help of a input output you can control the you can give the output. So in this way we have the see this. So with simple KVL KCL with the help of simple simple KVL KCL method we can convert into transfer function. So this is your transfer function. This is a first order system. When the blood from chamber number one to chamber number two this is a first order system and if we combine all it will be fourth order system. See this is the chamber means this is a capacitor consisting of the heart vessel and then this is a resistance and in this way we can simple KVL KCL method simple KVL KCL blood flow level with the help of simple KBL KCL method. So this is one. Now it is if you are considering chamber number both chamber one chamber two and output and output from chamber number system respirator when the blood is going toward the respiratory system. And this way this is a 70 ml put input 70 ml output must be 70 ml.

[00:51:45]they say and validate this result with help of a block diagram and the transfer function. So this is your mathematical model. This is your transfer function.

[00:51:54]The second resistor where R2 is the resistance of R1 R2 resistance of walls. C1 C2 the capaciting element of chamber one and chamber number two. So in this way you can convert into mathematical model and this mathematical model if you have validated your result with the help of simulation you can go for.

[00:52:14]So you can create your own systems.

[00:52:17]Chamber number one, chamber number two, chamber number three, chamber number four, chamber number three will have a one motor. Chamber number three have a second motor because they are against the gravity.

[00:52:27]This is a all this blood is going from the lower part of your superior inferior minava upper part. So this is the calf muscle will the will give you the strength. It will pump the blood whatever we have in the veins. So this will be pumping. So it is just like a second half. This is just like a second half. It will transfer the blood to the chamber number one against the gravity. So you can see think about this. This just like a pump. So this calf will give you the blood flow and this blood is deoxy blood is going towards the your chamber number one. So come to the fal monitoring. So why we are interested in fomemental monitoring?

[00:53:10]In fameal monitoring you can think this is normal heart and you have seen the normal heart the blood from the chamber number one then it will transfer to the chamber number two through this particular wall and this from RV to chamber number three from this particular wall going toward the respiratory system. The respiratory system will give you the good blood will enter the chamber number three left artera.

[00:53:35]Then from this particular wall it will be go to the left ventricle from this left ventricle to this particular aotic wall will go to the aot. So this a point and these points are what?

[00:53:49]Blood pressure is when the blood output at the time of starting of the this blood from this electrical electrical will give you the six blood pressure will be there and this signal will be transferred to the brain the controller. If there is increase or decrease in the blood pressure the blood the brain will give you the sympathetic signal and parasympathetic signal for controlling the heart rate and the blood pressure this inside the human body. See now our main component is the blood circulation of the fetus because fetus heart is not fully developed because means the pumping is not fully developed because why pumping is not fully developed because of your the lungs are not fully developed in the birth. So they if lungs are not working so we are not means it is not being inhaled then it will not be able to look. So everything is through the placenta through mother. So this side this is attached to the mother side the temporary organ the placenta just like a filter and the blood from the umbilical cord.

[00:55:02]This is code umbilical cord double connecting to the navl of this fetus.

[00:55:09]And you can see this is a umbilical arteries. Arteries are meant for good blood but here arteries are for deoxxy blood and veins are meant for deoxxy blood but here is for good blood.

[00:55:24]So the good blood will go to the fetus through vein.

[00:55:30]There is a one vein and two arteries.

[00:55:34]Three known as three blood vessels.

[00:55:38]So this is a normal blood vessel for a fetus. And there there are multiple fetus it will be having means according to that there will be a code. It may be one, it may be two, it may be three also. For three fetus, two fetus two, one fus one. For two fers it may be one also. And it is then it is anomaly. So we have to take care of that through the help of a fetus through the help of a mathematical model because the blood will transfer to the fetus through this particular vein. This way and it will be given to the complete human body the fetus sorry fetus and this the growth of the fetus depends upon the the blood flow from the fetus from the mother side through placenta. When the blood supply is very good, it will consume the oxygen and the waste product. The waste product will be transferred to the mother side again through the placenta with the help of a umbilical arteries. So it is known as three blood vessel. When we measure through ultrasound, if one blood one arteries are missing, what we'll have that is a critical condition. So again and again we have not to measure the ultrasound through ultrasound because if it is validated by the ultrasound okay there is a one artery one this artery is missing then why to again go for uh you can say the monitoring through the ultrasound machine again for this particular we can have the good mathematical model and this mathematical model through the predictive model we can create a predictive model and we can validate our result through simulation and then an alternative because ultrasound is not good every time we have not to use this ultrasound machine.

[00:57:30]So we should think about alternating machine also. So so this is your fetus.

[00:57:36]So I'll this is through mother side.

[00:57:38]This is your fetus. So everything will be taken care by the this placenta mother side. This is temporary organ temporary. There are various blood vessels but for the time being we have I have taken this the one vein for carrying the good blood and two arteries for deoxided blood waste product and good blood. So in this way you can simulate this result.

[00:58:05]So this particular before birth your this heart is not working but it it needs blood supply and this blood supply will be taken care by the placenta or through mother's side. This is through when before birth. See this is after birth it is closed. You can say this is closed.

[00:58:24]This is permanently fibrous mean it will be closed permanently. If it is not closed, it is known as heartto.

[00:58:33]When it is not fully closed, it is known as heart. So heart means we can create a mathematical model. When the input blood is 70 ml, output blood must be 70 ml.

[00:58:42]But because of this particular hole, it is lost. If it is lost, you will not get the good blood. And if you have not good blood, then it will disturb the internal organ system. They will get the less oxygen. If they will get less oxygen they will mean your cell will will die out. Okay. So this is your system. Now this is why saying open before birth this is newborn baby R newborn R. Okay.

[00:59:13]So through this particular system see this particular system. So this placenta and this fetus this can be converted into a mathematical model. This is the adult already I will discuss adult system because the lungs are working everything is every organ is working and you are capable of controlling your homeostasis and it will maintain adult system and it is fetus because fetus you can say the oxygen source placenta adult lungs bypass structure fetus uses three stunts adult no functional stance pulmonary circulation fetus minimum blood flow So adults complete polypary circulation means inhale that is pulary means external respiration that is pulmonary circulation means your respiratory system is also working the liver function fus partially bypass adult it is fully functional so this is a complete system and now come to this particular you think about it's a mathematical model so you using the markup model we can predict so you can say it is through mother it is mother side it is fer side it is node one it is node And node one is related to the mother's side. The mother will have some disease and that will affect the they will affect the blood flow. And to for the fetus there are some problem and that we will add. So with the help of a mother's side problem and the fetus side problem this we can we can make a model and then we can predict.

[01:00:46]So when we go for mathematical model you can this this is placenta towards mother's side towards fetus side and this is a signal flow graph it is a blood from the mother to fetus through vein otherwise vein is meant for deact blood but it is for carrying the good blood because the there are so many walls they can control the flow of the blood and in arteries it is able means from the fetus side the blood is going toward the so When it is being monitored by the ultrasound through ultrasound this one is missing.

[01:01:23]There may be one one this one is missing. So load will be on this. So why to monitor again and again through ultrasound machine. So we should go for some mathematical predictive models. So this is the PT with the help of uh this P here the blood this blood mean good blood is going towards the fetus. how much time it will take towards the fetus and how much is the quantity. Now P1T is the blood here deoxxy blood will go to the mother side and and based on that this is based on artery 1 RT2 and it is based on the best blood vessel of the fetus or you can say the umbilical c. So this is your PT the output blood when it is being converted into a simple equation it is a PT. PT means the blood flow velocity or the characteristics at node one. Node one means towards the mother side and P1T P1T means the probability of occurrence or probability of flow of the blood or velocity from fetus to the mother side. And this is for a for a you can say the best combination of the fetus means all blood vessels are intact in the fetus and this is for a one fetus it may be two two fetus then duplicate of this three fetus then triplet and if let's say two fetus and there one blood vessels or one umblical code then the what is the load on this particular the pathway you can say like this so this is a mathematical model to mark model it's a predictive model. We can predict this state from the input side and from this this is input side and we can predict this state also. So it's a transfer of P 0t from mother to fetus.

[01:03:15]So probability how it will go to the fetus and this is how from the fetus to the mother. If you simulate this result uh okay before that I will see the factor affecting the node one high blood pressure to the mother diabetic to the mother infection to the mother kidney disease heart or respiratory disease alcohol or drugs so these are these parameter these parameter towards the mother's side or towards the node one and so this is a just five minutes just five I will take just five minutes Okay then I will stop talking. Can I take 5 minutes?

[01:03:56]Okay. So factor affecting the maybe two blood vessel long blood long code long code means it if it is long code it will be a tie in the the neck area and it just like a fy the local code if it is node is very less short or local code there is node it will reduce the blood supply from node one to node two node to node node one and the small.

[01:04:19]Okay. Now this is your if you go for simulation you go this is your the result from the ultrasound this result from the result from the blood flow and average value this simulation result. So both are same so we can define so predictive model. So if we have a predictive model we can reduce the blood hypox fetal hypoxia oxygen oxy carbon dioxide hypoxia fetal death low birth rate and presenter we can go. So this is a snaps I will skip this. This is the you can use this combination in your neural network.

[01:04:52]This you can use for neural network because this is your neural network and this neural network will be connect to snaps the neurons dendrites cell body and snap. So this uh okay this is a validation this is for Thank you sir. Thank you very much. Any question any doubt you can ask. Thank you. Your time is very less otherwise I would have explained better. So if any and you want disc more discussion I am available also anytime after the conference also. Thank you.

[01:05:38]Okay. Should I should I close?

[01:05:53]>> Your sound is not clear.

[01:06:02]>> You can ask any questions. Any doubt?

[01:06:20]You can use this.

[01:06:37]Good morning.

[01:06:43]Okay.

[01:06:56]Sir I have one doubt.

[01:06:58]>> Yeah.

[01:06:58]>> So as you know that all physical systems have to be uh before we use it for AI or neural network the first job is to build a mathematical model.

[01:07:09]>> Yeah.

[01:07:10]>> And we need to give an uh hypothetical data.

[01:07:16]The data is to be actually collected in the physical system before we supposed to train the network. Is it right?

[01:07:23]>> Yes. Yes. Yes. See whatever that is I have said.

[01:07:30]Yes.

[01:07:32]You please carry on sir. Go ahead.

[01:07:48]Uh hello sir. Am I audible?

[01:07:52]Yes. Yes. Yes.

[01:07:57]Yes.

[01:08:00]Sir, am I audible?

[01:08:02]>> Yes. Yes.

[01:08:05]>> Yes. Please, sir.

[01:08:06]>> Sir, am I audible?

[01:08:08]>> Yeah. Yeah, you are audible.

[01:08:10]>> So any before we use it for AI or neural network or anything, we need to convert the physical model their parameter We been supposed to build a mathematical model first.

[01:08:25]>> Yes sir. Bill. Okay. Yes. First mathematical model.

[01:08:28]>> Then the network is supposed to be trained >> using the actual experimental data.

[01:08:34]>> Yes.

[01:08:35]>> Then the network is to be trained.

[01:08:38]>> Yes.

[01:08:39]>> That will predict the uh other parameters.

[01:08:43]>> How do you use the network I mean AI in this in this situation? It is a it is a mathematical model. Then through mathematical model phantom study.

[01:08:55]Phantom means hypothetical model of a physical system. Just like just wait when we are because by working as a uterus papaya is just uterus that is a phantom for uterus.

[01:09:09]So through through through papaya we can we can discuss the or we can validate the phantom or we can validate the utus of the pet system in mother and then we should go for clinical trials that we have to take ethical clearance expected rate of acceptance of the error.

[01:09:33]Pardon?

[01:09:37]what is the expected uh error how will occur in by using the AI tool?

[01:09:43]See in AI system so uh you just I have shown you the slide and predicted data what is the error expected during the study which is allowable error to be permitted >> because because I have used impmography and because to validate our my my work and in that case we have to go for more and more recyclical processing and image recction algorithm for better better response. You can say like this.

[01:10:52]on behalf of >> on behalf of ATA Institute of Engineering and Management Sciences, Benuru and the organizing committee of the International Conference on Trustworthy AI for Intelligent Computing Systems 2026. I sincerely thank our estin keynote speaker Dr. Shin Kumal Paka for an insightful and engaging session on mathematical modeling physiological systems. Your exams in 2v2 metab monitoring showed us how can truly impact health care. We are sincerely very grateful to you sir for your valuable time and expertise. Thank you once again. Thank you madam. Thank you.

[01:11:33]Thank once again. Thank you everybody.

[01:11:36]Thanks for listening. Okay. Thank you sir. Should I leave?

[01:11:44]Yes sir. Thank you. Okay. Yes, sir.

[01:11:48]Thank you.

[01:12:59]Hello. Hello.

[01:13:02]Hello. Good morning everybody.

[01:13:05]Uh the participants who are physically present in the auditorium as well as the participants who are in the online. So there is announcement of best papers.

[01:13:16]Okay. which uh the different tracks were held yesterday. Okay. We have around 10 papers.

[01:13:24]We have around 10 papers best paper awardies. So we have a validtory function in the afternoon at 2:30 a.m. in the uh aditorium of Amrita Institute of Engineering and Management Sciences Beri.

[01:13:39]Okay. These 10 papers participants I will announce and if you are interested you can come out to our institute and collect your best paper award certificates. Okay. The best first best paper award goes to uh T2064 engineers desk a web-based solution. The authors connected to this paper. You can come at 2:30 to Amrata Institute of Engineering and Management Sciences Bed today. And the second paper T10067 e-learning platform using AI.

[01:14:16]The third paper T4003 design of IoT based medication.

[01:14:22]Paper number four T10025 road safety.

[01:14:27]Paper number five. T4078 UAM based functional verification.

[01:14:35]The next sixth paper best paper award T2043 realtime object detection.

[01:14:42]The seventh paper T2049 IoT integrated smart energy harvesting and automated irrigation system for the sustainable agriculture.

[01:14:54]The eight papers T1015 clock design generator using machine learning algorithm. The ninth best paper T1060 IoT based LPZ leakage detection.

[01:15:10]The 10th best paper award T4046 enhance the efficiency of the thin crystalline silicon solar cell through advanced light management using aluminum.

[01:15:23]Yeahsuming Al23 it seems nanop particles. Okay, these are the 10 best papers awarded yesterday. You can come to Amruta Institute at 2:30 a.m. today. Uh 2:30 p.m. today. Okay, to get the best paper certificates. Thank you all.

[01:15:52]Now you have a tea break of 10 minutes.

[01:15:54]After that at 10:30 I request all the participants to present at the venues what we have displayed in the projector.

[01:16:02]Thank you one and all.

[01:38:27]myself Dr. Kumar B professor and head for information science department as college u and uh with me today's uh chair session chair is Dr. uh T son Kumar sir he is head research and R&D at ACS college of engineering and I heartly welcome on behalf of our college to you sir thank you sir we'll start the session sir so first paper uh t012 uh sorry 021 paper number 021 please come and the next paper is 29 series T1 series 29 be ready with the paper presentations Please.

[01:40:44]Camera. Smile. Smile.

[01:41:02]Hello.

[01:41:13]So good morning everyone. My name is Sita. So today I'm presenting my paper on a multimodel multi-agent healthcare conversational AI system for clinical decision support.

[01:41:33]Hello.

[01:41:40]So I'm from BGS Institute of Technology, Manda.

[01:41:44]So these are the contents uh I will be walking you through.

[01:41:51]So introduction. So artificial intelligence in healthcare is crucial and is very uh accurate and should be precise. So uh today's large language models like OpenAI, Gemini and Chad GPT uh these are not domain specific models and also causes hallucinations. So multimodel AI uses uh users to interact through speech uh text and images. So it improves accessibility and users experience. So we have u uh introduced a new approach called as rag which is called retrieval based uh retrieval augmented generation. So it can produce accurate and useful information for the user.

[01:42:39]So problem statement. So first problem is hallucination in alms like chat GPD or Gemini.

[01:42:47]So what the what this models can cause is so it uh if you give uh user gives a query so what is the dosage of paracetamol so chd can say uh like 300 mg for uh any user. So it is not domain specific and it is trained on robust data. So we are implementing a retrieval based approach called as RAD retrieval augmented generation. So the model do not causes hallucinations.

[01:43:18]So the research problem is how we can develop a safe accurate and multimodal healthcare conversational AI system for clinical decision support.

[01:43:27]So object is we are developing a healthcare conversational AI system which supports text, voice and image inputs. So implement multi- aent architecture using retal augmented approach. So this uh helps in reducing hallucinations and improves factual accuracy.

[01:43:48]So this enhances uh health uh in healthcare by providing easier accessibility and usability.

[01:43:55]So these are the literature survey we have uh gone through. uh we have considered so many advantages of previous uh researchers and also uh we have uh improved the uh withdrawals uh so this has helped us to uh create a more domain specific website for healthcare support.

[01:44:14]So methodology is simple so first input stage a user user gives a query through voice or chat message. So next this voice is converted from speech to text through faster whisper. So faster whisper is a speechtoext conversion model. Next the knowledge retrieval. So the the rag uses uh retrieves relevant documents from the database. So the ll next generates this response into a simple natural language. Next the output is generated through uh text to speech which is a python library called htts.

[01:44:55]So next interfacing deployment we have deployed this um project in a cloud platform. So users can u users can interact with this uh healthcare website.

[01:45:07]So this is the system architecture or model.

[01:45:12]So first the user uh gives an input through voice text or image input. Next the voice input is converted from speech to text through faster whisper. This is the processing level where all the generation and retal augmented generation happens. So for local we have used colama or ma. So this is a hybrid approach. So the rag uses uh it retrieves relevant documents from the medical database. Next this text is converted into speech or if it's a first year query it is generated in animated video. Next all the databases are stored in a local or cloud database. Next uh the output will be uh animated video for first year guidance or if it related to any first year uh sorry drug information lookup this can be uh like chat inference. So all these are happened in a web interface layer which is deployed in streamllet using flask front end.

[01:46:16]So these are the results. So this is a hospital website. It contains three models. first guidance, appointment booking system and uh uh drug information lookup. So uh first first below uh there is a hospital appointment booking system. So users can talk to the AI model and can book appointments through speech to speech. Next is the first year guidance voice assistant. So the user can talk to the model and uh he can in at any emergency situations he can ask the model. So it can generate animated video support. Next drug information lookup system. A user can uh upload any uh drug related medicines through labels. So OCR uses these labels to convert into text and it generates dosage of the medicine.

[01:47:09]So conclusion developed a multimodel healthcare conversational AI system and we have implemented rack multi- aent to improve reliability and achieve safety in responses. We have enhanced accessibility through voice text and image inputs. So our future work uh future work uh consists of uh improving the medical database into a larger extent. So we are uh improving the cloud to develop into a robust u cloud model.

[01:47:39]Next uh we are seeking information from healthare professionals to validate the answers so that the model can give grounded answers. So these are the references we have referred. So thank you and questions are appreciated.

[01:48:03]Okay. The my first question uh you are using a based fine but uh the major problem in medical industry without professionals you are prescribing medicine is unethical.

[01:48:17]>> Sir that's what I said sir. No no first you understand >> without prescription without consultation prescription medicine is unethical because paracetamol you are telling what are the purpose we can use paracetamol do you aware of this okay you tell sir like uh I said now sir the documents are already verified by the healthare professionals >> no my question is if you are recommending paracetamol what are the purpose we can recommend paracetamol So >> what is the causes of paracetam?

[01:48:52]>> Yes sir. It will describe the causes also and it will also describe how how much dosage the user can user can take.

[01:48:59]>> That's why what is my point? Now in the entire system doctor role is missing.

[01:49:06]>> So no sir these are verified by doctor also >> where you can go to your output of that last uh where is the doctor consultation. These are these are already verified by the doctors.

[01:49:19]>> No, here there is no >> we have provided this disclaimer sir.

[01:49:23]These are we have provided disclaimer that these are only verified by doctors and these data sets are limited. So so it can be correct and uh we will provide the disclaimer. So this is an AI. So it is grown dead but >> now you understand the point uh point number one by seeing physically only the doctor can recommend the medicine. So this is not a recommendation website.

[01:49:50]This is only tells the dosage and >> that's why I'm telling see paracetamol we have 500 mg 650 mg.

[01:49:58]>> Yes sir.

[01:49:58]>> Two things are there. What is the causes of paracetamol? Be aware of this.

[01:50:03]>> So yes the website can give everything.

[01:50:06]>> No that's the problem. Actually website should not give doctors should give actually mean medical industry if you do such a kind of things we have to we have to get first ethical permission. See one thing help you I have some problem without consultation doctor I can go to pharmacist I'll tell this usually in India what is happening now I'll go to pharmacist so I have a cough I have a cold I have this problem the pharmacist will give medicine okay now it is happening in India okay the pharmacist will give the medicines but he won't take any accountability something is going wrong who will take accountability in your project that's why we have provide disclaimer sir this is not a >> no disclaimer is something is I lost my life I'm I'm using this system something is happening I lost my own organ for example if you are taking over storage >> this is charging >> it will affect kidney it will affect liver then if I lost or something is happening I'll go to some doctor doctor will ask to whom permission you have taken this medicine what answer I can tell to doctor I he said I will not treat because you are taking medicine without my permission. I will not trick and most important thing paracetaml I'm going to take okay already have so much existing disease who you take accountability sir this only provides information sir that charg will give yes sir it will give but it will hallucinate so this is not trained judg is trained on robust data including medical everything it is trained sir but this is literally trained on medical data >> how much data you have collected for this uh because you want to train sir How much data have collected?

[01:51:50]>> Very very limited data sir.

[01:51:52]>> How much data? I'm talking about size of the data.

[01:51:54]>> Uh like literally it is uh 1,000 uh where are you taking this thousand,000 data?

[01:52:00]>> Uh by first you will collect from medical uh sites sir.

[01:52:04]>> Okay. You are taking from is authenticated sir you have taken MDP8 sites?

[01:52:09]>> Uh yeah from who website is there sir which >> I'm talking about MDPA1 site is there for medical data. Have you taken contacted any MDPA sites?

[01:52:19]>> No sir, we have not.

[01:52:20]>> Then s if you are taking medical data who will authorize one doctor supposed to authorize >> sir first. So we have taken the data and then we have like in our college only there is hospital.

[01:52:32]>> See you you are not understanding the point. Every medical college there is called one ethical committee. I want to use such a kind of things. Now first I have to approach ethical committee. I have to present my project. The mythical committee is supposed to approval you know any trial I want to do in any of the places even if you take vaccination any medicine I want to do human trial I have to take ethical I'm telling you are simply recommending paracetamol 500 or 650 I have already kidney issues or already have liver issues if I'm taking paracetamol continuously it is indexing of my kidney problem as well as liver problem then who will take accountability they'll directly put case against you only >> sir that's what that's what I said sir >> disclaimer will not >> this has happened on charity sir this is open has happened >> your system also will happen like that only >> sir by providing the disclaimers and specific guidelines >> disclaimer is just one message when I I'm directly using your system I'm taking medicines based on your recommendation my body will affect who will take accountability whether >> yes sir user is user is user takes accountability Because you said no sir wait sir wait sir I will I will tell >> yeah one one I will add to s user is the responsible then why why you will take this kind of why we want this kind of specific I tell you one thing if I'm saying by YouTube now YouTube will not take responsibility but your system is I'm enrolling your system then you have to take responsibility you are not enrolling the system is not enrolling the system >> then what I'm going to do if you don't develop them if you are not taking responsibility why we why you want to give this to the societ Sir, no sir. This is only for research purpose and this researches are for good things to do some good things to the society right no application purpose get in India you understand the first system for how the medical is functioning anything I want to do human trial is supposed to get approved from ICMR.

[01:55:03]>> Yes sir. All the medicines you should get permission from ICMR and ICMR they putting clear information any human trial they want to do as a human first I have to enroll myself I have to give self declaration yes any causes will happen I'm taking my own responsibility that I have to register in ICMR then they start human trial okay they have to analyze periodically it is not easy because most of the thing why A is not coming to medical you know what is the reason >> that is the reason sir what is the reason >> two departments still they not get into a one department is medical industry what is the second one you aware of this law sir >> law industry they don't accept because these two are sensitive issue still now they not allowing even supreme court also because I'm telling one other way instead of this I can develop a based lawyer so the a will going to appear on behalf of me it will counterfail okay the problem is the a lawyer fails who will take accountability that's why it's more sensitive simply I can develop one system I can uh simply post it then any media or you can go for websites now directly they'll put case against you only you know ICMR national council and Indian medical association all three they'll put case against you they'll put 1,000 2,000 3,000 crores defamation suit against you because he is giving wrong things. Already there is a problem homeodi alopodi sittitha natural bodhi and ambs these mbs people not accepting homebody alopi sitta natural all x y z medicines there is a contradictions going on be aware of this s home people should not you recommend alopuri doctors still the problem is exist that problem government not able to solve this is I think maybe your intention is good but it is very dangerous federal system. This is not about uh implementing the health care sir. This is about >> first one, second one at least first aid somewhat. Okay. Drug is danger. Yes, I get it sir. I get it.

[01:57:22]>> Drug is too much danger.

[01:57:24]My family doctor will say I should not take paracetamol. Your system says go and take paracetamol. It will not say don't take sir. It will it will first describe how to use paracetamol. But you tell how to use paracetamol. Sir I I don't I'm not you tell how to use paracetamol. So I will first describe what is paracetamol and how is it used and what is it used for? I'm telling you you developed the system you developed the system which medicine you have used drug information paracetam you only use paracetamol. Now how to use paracetamol?

[01:57:56]>> So it is it is uh described by >> how much dosage I have to take paracetamol. It will say sir but in last it will say this is not a um um medical prescription by doctor this is completely agend prescription he will take accountability I can go to court I can show the evidence this is prescribed by doctor I have problem because of this doctor a generate now who will take accountability that is that is why we are repeatly asking accountability is missing >> that is where ethics of AI come here sir It fixes with so much field there but medical is too dangerous.

[01:58:37]That's why actually please do something you have to do ground study and consult with the doctor. If you have any doctor you can consult you can present to the doctor then you can get opinion of the doctor that is most important.

[01:58:50]>> That's why sir each and one medicine is done by the doctor. So that's why our database is very limited. Okay.

[01:58:55]>> Okay. Thank you. Thank you.

[01:58:57]>> Thank you.

[01:58:58]>> Take care.

[01:59:00]Okay sir, I will take that as a suggestion and uh I will try to implement.

[01:59:03]>> Okay.

[01:59:03]>> Thank you sir.

[01:59:04]>> Next paper uh T1 uh 29.

[01:59:17]So paper is there anybody paper number paper ID T129 and next is T134.

[01:59:36]So paper ID IC TA ICS 2026 T134.

[01:59:42]Yeah, please come. And next paper will be uh T137.

[02:01:37]Good morning to one and all present here our esteemed chairperson and all our beloved faculty members and my fellow pass participants. I am Sujana Desh Pande from information science of sixth semester. So uh I'm representing uh my paper title as God's eye an AIdriven multisource intelligence system for realtime human tracking and predictive analysis. My team members are Simon Le Alexander Shyavadi and Shibramma R under the guidance of Dr. Kumar Di.

[02:02:07]So these are uh these are the top uh out this is the outline of my presentation which will be covered today. First will be the introduction and the problem statement, objectives and lit literature review, proposed methodology, system architecture and algorithm and topology used at the last with results and results and conclusion and the feature scope moving forward. So the challenge in today's uh modern surveillance system is in digital era surveillance uh system generates enormous of volumes. We're using smartphones and other GPS technologies. So this uh this traditional surveillance system mainly operates independently and lack mult uh lack intelligence required for realtime uh analysis multiple source integration and automated human technology and predictive monitoring. So our solution to this is God's eye an AIdriven intellusion surveillance system which provides realtime surveillance tracking movement across frames and integrating multiple data sources plan prediction for of future moments.

[02:03:13]So problem state uh our problem statement is traditional CCTV based systems suffer from uh several major limitations that reduce their predictive effective uh effectiveness in real time environment. So manual uh traditional surveillance system mainly depend on manual dependence manual monitoring dependency.

[02:03:34]Nor normal CCTV surveillance system use normal manual monitoring which needs labor uh labor effect also people man people need to sit manually and monitor the CCTV surveillance. Our next limitation in the current surveillance system is lack of intelligent detection.

[02:03:50]There is no smart detection in our present surveillance system like there is no automatic detection of the uh activities occur occurring at a particular location. So next is limited track tracking capability. Each our present surveillation system in uh CC cameras or any other surveillance tracks works independently due to which it cannot club with other frames or other surveillance areas and cannot track the location of the person of the person or the human at the and the future moments or the future locations. Next, absence of predictive analysis. So uh normally our CC cameras or surveillance have absence of predictive analysis like they don't predict what the future what is the by the behavior like what is the future moment of the person or the how how is it how is he going to uh move going to be moved forward in other locations or he's uh track the person is a convict or a person who is wanted in our criminal record etc. So next is isolated data sources. Most uh systems process video uh feeds independently which makes uh which makes which is a major reason we don't have uh continuous surveillance. Uh if if the systems are not linked and they're independently processing. So if there is a particular event occurred in a location and if he moves to the other camera location we don't get proper tag. For that we need a manual intervention from once if the person is uh uh detecting in one uh seeing the camera in one position. So after seeing that he have to he has to go to other camera position and check the CCTV footage of that particular location. So you cannot see the frames at together to unified platform. These are the limitations which we have in our current uh modern system.

[02:05:41]So due to this we have an impact of delayed response and uh during emergencies reduce surveillance efficiencies and increased risk of mis mis miss missing critical events due to human interventions. Sometimes we miss critical evidences which may lead to many other pro uh problematic stuff like we miss evidences for most of our uh uh like criminal activities which will lead to uh many other consequences.

[02:06:09]So objective of our project is to improve human tracking system. It is you uh so first objective is is to improve human tracking system in which we develop an AI system which is capable of accurately tracking individual in real time. So next is enabling smart surveillance system. We uh we are using camera IP networks, camera IP networks and capturing all the uh all the nearest camera surveillances making them to assemble in a unifi uniformed platform for implementing the surveillance which helps in um getting all the datas from the sources near to the uh uh near to that location and getting the track of evidence where the uh exit convict is moving around. So next is uh support uh supporting missing person identification. For example, if a person is missing from a particular spot and there is a CC camera which has captured that the person has been kidnapped or he's missing from that spot. So if we track uh if we have this type of uh system over there. So what will happen if the person is captured in that camera. So he'll be if the person if the van carrying the person moves some on the other location by contacting that other location camera we can access that person we can track that person immediately and that person uh that missing case report will be immediately sent to the police station or the crime branch for the further investigation.

[02:07:34]So next is uh provide predictive insights, analyze movement patterns and predict possible feature location using machine learning techniques. So by these using uh by these cameras uh this uh professional AI intelligence system we can have uh predictive analysis like where the person will be moving further or the way the person exact location is predicted at uh by using these camera uh camera frames. So next is enhancing disaster uh response. So uh normally during disaster response it is very difficult for the team to search for the people who are uh injured in the disaster systems. So by using this system uh we can track where are the where are the more uh people getting have got injured or the people who have got stuck in the disaster or after uh after the disaster is completed the people can go rescue the people immediately. So this will help in saving the people earliest. So next is promote ethical AI usage. Uh with this uh we can uh use this majorly project can be used for many cu uh criminal services and intelligent intelligence department of India to find the uh convict of the terrorist or the other part of other criminal activities happening in India.

[02:08:51]So our main foundation for our project uh by which we have got idea and what are the limitations which we will be know we have had been known by this literature reviews. So uh SK SK sing in RK Bispas uh 2017 this paper was published. So this is this was only contributing about multi- camera human detection. There was no predictive analysis in this there was a limitation.

[02:09:19]Then a WHMA and uh AK Singh which was produced in 2019. This is a realtime detection using CNN. So but this but there was a limitation which was single source processing only one side the processing was happening like only what at a particular location it was happening. So uh Satya and Darthik this paper was these were the authors where this paper was published in 2020. This was for introducing surveillance system but the major uh limitation was there is no multimodel integration like there were no multi- camera integration to this. So next is uh pira shaker and ramen in this paper was published in 2021 person reidentification using DNN neural network. So limit uh this is there was limited uh limitation is visual tracking there was no visual tracking in this model. So Shitza and uh SK Sync again in 2020. This is trajectory prediction model uh using ML. So they uh they used to uh in this model they used to uh predict the uh pedestrian walking. This was prediction uh pedestrian tracking uh used in this. So there was no uh limitation to this was there was no realtime multissource fusion. So this was the research gap which we uh these all are being trying to tried to implement in our model. So existing systems mainly focused on detection, tracking and prediction individually. So very few system provide unified AI framework realtime prediction and multisource intelligence. So the proposed guard god system brids the gap through integrating intelligent surveillance system in one system.

[02:11:03]So methodology we have used agile plus iterative prototyping development methodology in which uh first step is data collection. We collect the input from various CCTV cameras, webcams and other feeds of uh the resource which are integrated normally in our public sectors or any other uh development units like uh school, colleges and all.

[02:11:25]So next is pre-processing in which frames are normalized and data is been prep data preparation happens. So next step is human detection. We have used YOLO V8 model for human detection and real-time object detection which gives a bounding box to the particular uh human like if you enter a data person's data.

[02:11:43]So it will be tracking the person wherever it is. If he comes under the camera surveillance there will be a bounding box with ID and the name of the person which will be showed in the u showed in the CCTV footage. So which will help us to track wherever the person moves. So next is tracking. We use deep sort algorithm for uh tracking.

[02:12:05]Person will be given as I said person will be tracked first. You'll be given a unique ID assessment and continuous movement analysis will be done using the uh deep sort algorithm. So next is prediction engine. We use future trajectories and possible movement patterns like where where the person will might move further. Will he be going this way or will he be going this way will be predicted like uh predicted early. So next is visualization realtime monitoring dashboard will be there. So in that dashboard we'll be monitoring the person uh fre constantly and uh tracking the overlays wherever he goes he'll be tracked.

[02:12:43]So this is our system architecture or the model which we will be using to implement. So first is uh data injection layer or data acquisition layer where the comp uh where it collects uh information from inputs from camera or the GPS sources and the image data sets which will be provided. Next is pre-processing layer here will be the frames will be enhanced or the images which will be provided will be enhanced and the data will be normalized. Next AI core engine that is the core engine for the detection is YOLO 8 for detection and deep sort for tracking.

[02:13:19]Next is fusion layer. It combines visual data, location information and tracking outputs. Fusion layer is a layer in which the visual uh visualized data the major data will be uh seen and the all the tracking information will be stored there. Next is prediction engine where moment of analysis and future trajectory prediction like future the way the person will be moving will be predicted here and the visualization the final layer in which bounding boxes that is if a person has been f the person's data is found in a particular area that bound that person will be surrounded with a bounding box which will uh be given a unique ID and a name so that wherever he moves that person's uh person's tracking will be continuously done results. So currently uh our system uh in our system realtime human detection is achieved successfully in which it processing speed is appro 25 to 30 frame per second. So accurate person tracking across video frames whenever the in the CCTV footages when the videos are moving uh accurate person tracking is happening. So stable detection under varying light conditions and bounding boxes are generated in real time when the person's data is in uh implemented.

[02:14:38]Unique ids assigned for the continuous tracking of the person. So which are key achievements are improved surveillance accur uh accuracy. AI based detection is significantly monitoring the efficiency.

[02:14:50]With AI we can significantly track the person wherever he is moving. So reduce human dependency. So normally in our previous surveillance system there was much dependency on human because if anything happens for the CC camera they have to go to the uh particular monitor they have to sit and rewind and watch everything again and again. So with this help of this uh project uh the surveillance system we there is no much efforts of human intervent. So if uh the moni uh if we implement this the person will be tracked continuously and if is near to some uh places immediately the alert will be sent to the particular police station. So uh enhanced intelligence detection is h detection tracking and prediction uh with unified framework in one framework all the all the things will be happening which will easier the human uh effort and will be giving uh uh good surveillance for the compared to present surveillance systems. So uh scalability potential the architecture supports future expansion in multi- camera environments smart city applications and uh data disaster response systems.

[02:16:00]So conclusion uh the proposed broadside system presents an AI powered intelligence frame uh surveillance framework capable of which is realtime human detection, automated tracking and multissource data integration and predictive movement analytics. These are implemented successfully. The implementation successfully demonstrates the all of the following uh points. The system establishes a strong foundation for surveillance and predictive analysis applications. So fcop is um further which will be implemented is multi- camera integration. Uh so after uh multi camera interaction and facial recognization face uh by using deep face or faceet algorithms we can implement facial recognization of the uh person for tracking efficiency and smart alert system. So if the suspicious and convicts are near to our boundary or near to the position which is marked uh we can get generate a smart alert system which will be sent to the nearest uh police stations or the crime branches.

[02:17:02]Next is uh web dashboard realtime monitoring dashboard in which uh we have uh currently we have provided the features like uh priority based uh con uh human track uh detection like if we are given higher priority priorities of for the details of such a person the person if he's sounding to the surveillance of that camera he'll be detected immediately and we'll be sent information to the nearest crime range.

[02:17:27]So next is uh disaster disaster management in integration which supports rescue operation through the crowd when the disaster has occurred.

[02:17:37]Cloud and edge development which improves scalability and remote accessibility. With the help of cloud and edge development we can access uh we can implement the surveillance system in the remote areas which will help uh good tracking system in the remote areas which was normally uh less accessible.

[02:17:55]These are the references or the foundations which help us to get uh get more information about this uh topic in the uh world.

[02:18:06]Thank you. I open to any questions.

[02:18:19]So you have done this project or still it is in elementary stage. Actually sir this is our major final year project which is being implemented uh in the first phase uh in which we have implemented uh >> in first phase which are the things you have implemented.

[02:18:33]>> Uh we have implemented a realtime uh detection of the person bounding boxes will be appearing in it. So it will be given generating a unique ID. Uh we have made a dashboard in which uh we have given a target uh addition. So if you have the data of the target the target will be added into that uh web page and we'll be set with the priority.

[02:18:54]>> So this project have you done for your college or general public?

[02:18:59]>> Uh both sir it is for general >> in your college how many students are there?

[02:19:04]Approximately 2,000 students are there?

[02:19:07]Um maybe sir I >> um >> diagnostic tools are not available to diagnose this learning u lesser screening availabilities lack of open access and data because this is a very confidential data schools do not prefer to share the data of the students. So there's a lot of uh you know um disparity between the data which is available. Then we come to multilingual data sets. Uh you know uh according to the census of India I think we have 122 major languages. So a lot of multilingual uh data sets is also missing in this case. Um this is how we started our journey with the biblioraphic analysis. I mean we saw that you know the computational models were started building in the year 2010 and this is how the research is going so far about the scientific production of learning difficulties and using computational models and specifically AI. Uh if you see this graph I mean the the darker the graph is you will get to know that how India basically is a hub of the research and collaboration over here. For instance, India appears as a central hub with connections to other countries like USA, European countries and Asian countries. The dark blue color of India appears that it's a significant hub of research and collaboration with multiple countries.

[02:20:31]uh this is a graph that I have taken for the continents and you can see that the in in different continents it's higher in North America that's the diagnosis is between 8.5 to 23 percentage of school growing children while in Asia it is still lesser which is from 6 to 5% of schoolgoing childrens uh this graph represents how things are basically changing in this diagnosis and mostly China and India basically are the hub for the research in neurode development sciences and in these difficulties as well. Um so these are the things that we basically did uh in our research to find out who are the authors basically who are working on this and also we found out the that the India China and USA basically works heavily in this uh so I'll just skip all these uh the occurrences of so you see the machine learning and all these researches that we did appeared maximum then uh SP machine learning and all these Uh so this is our literature review. We started with 260 research paper that we got uh you know this query that we ran.

[02:21:44]Now uh this is all between 2010.

[02:21:48]So the research basically is growing in this field to 34.29%.

[02:21:54]Now yeah this is our literature review where we reviewed around 490 research papers and based on you know all the uh ifs and buts basically we came to 74 eligible research papers for our research. Now I'm not going into this but this is how the literature review basically worked for us. We found the theme we found the accuracy as well but there's one thing which I really want to uh point out over here. If you see most of the researchers that the accuracy is questionable and also they only work on very limited features. Maybe someone might be working on MRI, NLP and all these things. So these are the latest research papers which are but again there is a problem with these research papers. If you see the yellow highlighted ones, one research paper comprehensively work on the reading skills. One basically works on handwriting and one basically will you know work on the Prisma basically that is how the transformation takes place. So all these things are limited.

[02:22:54]So our aim basically is to build a multimodel that works really well in this case. So these all are the researches that we basically selected.

[02:23:02]So this these are the literature gaps that we found in this study. So you know uh there's a single modality approach that means we are focusing only on a single approach when we are doing the research. Um we there's no multilingual or the diverse data sets which are available like maybe you know the data sets of north India, south India and the other parts of country might be a bit different. There is a problem with the longitudinal study. So this dyslexic children they require time to be monitored. So when we say longitudinal studies we have to monitor them for one year at least in order to find out what kind of uh you know problem they are facing in the divergence. So and mostly there is a problem with the AI model because they get so much of data but the fact is you know uh there's no clustering there's no you know balance kind of a data that basically comes in from there and it's very generalized also.

[02:23:56]So uh the tools which are available they are very expensive in order to if you go for a dyslexia test or to a cleaning or to a special education need uh educator it's very weak and they're alignment with the current uh you know the UN and NP policies are also missing and most of the online tests which are available for uh this dyslexia are basically very English centric and that is not suitable for a lot of multilingual learners. Now this was our aim basically uh you know when we started uh building this tool we found out that everything in this plays an important role uh you know starting from the family history of the diagnosis then the phonological awareness the reading examinations the working memory trials and the intelligence test as well so uh this is the diagram that I want to show you this is how basically we came with our proposed model that we will be using a multimodel data collection over here that will be collected Think all these things the student handwriting.

[02:24:56]I'll tell you a very basic thing that happens in handwriting which is a student which is dyslexic might write B or D in a in your very opposite way.

[02:25:06]There is a problem with the other things as well. There might be a problem with the speech. There might be a problem with the eye movement. The retina movements is very limited. And also we tried checking the brain mapping using EEG. EEG stands for electrons fellowraphy. So that basically gave an idea of how basically the brain mapping is done. Then we did data prep-processing the noise balancing. It works like a data warehousing where we did you know extraction transformation and loading. Then we will be using some deep learning architectures. with her.

[02:25:38]There is something that we want to incorporate which is explainable AI that basically gives you more interpretation, transparency and uh the two models that we evaluated for explanable AI where the line model and the chappley model. Uh this is a longitudinal evaluation. So children will have to undergo for a test for nearly one or two years. And finally there can be a tool which can be available for cloud and u you know offline or online uh versions as well.

[02:26:06]So this is how basically we did the um implementation and the methodology. We did the literature survey. We have started collecting the you know development and the collection of multimodel things and we are basically making some kind of a model. We are also making a game that can basically help you uh track your retina movements instead of EEG. So uh this is how basically it works. Uh so we have multiple school age children who are participating in right now. We also collaborating with the NOS's to get their data of neurody diversified childrens. Um we are taking multiple things again from handwriting to speech to eye tracking to EEG and multiple cognitive assessments as well. We are trying to make a standardized psychometric test. We are trying to make an EI model which is called cognify.ai.

[02:26:55]I'll show you after this uh if that will be permitted. And uh basically the implications I mean with a multimodel there can be a procedure which can be 95% and above there can be personal recommendation systems that can be made for the client as well and this definitely is going to enhance the clinical and the research practice at least.

[02:27:16]>> Good morning everyone I'm hersa and uh this is my uh presentation I mean this is my paper which I've been working on this lately. uh these are uh AI voice agent based interview preparation platform uh which is which is under the guidance of uh professor Shinwasa and uh these are all the teammates uh I'm the uh I'm the team leader and uh these are my teammates and uh coming to the introduction uh 73% of our job seekers report that lack of the interview preparation and practice is the primary reason primary reason for the rejection the problem actually exists the thing is that millions millions of the students will be graduating with the right skills yet they fail to uh crack the interview.

[02:28:00]The main thing is that they'll yeah uh the millions of the students will be graduating with the right skills but they uh fail due to the right uh uh preparation and learning path and this is the reason we have uh made this web platform for this thing for tackling these issues and all and what this addresses is that uh uh students will be having the right skills and they can annunciate and they'll be they'll communicate with the skills but the thing is that they can't speak about what they have been working on. Uh they have been not thought what they have to speak about the skills and what they have worked on. The traditional interview platforms exist but either they will be costly or they'll be not having biased fairness and all things.

[02:28:47]So we have a traditional interview platforms methods which lacks the realtime interview uh interaction and all and adopt your uh evolution and performance analytics. uh it will be either very costly or expensive where the normal student can't uh uh can't afford that and it will be very biased for that. So it will be having some unfair for that and it will be very uh uh what do you say your P will help you to make you comfort for your evaluation and all. So that's the reason we have developed a seat which is an AI voice agent based interview preparation platform. It is designed to stimulate the realistic technical and nontechnical interviews and all. The system is integrated with voice agent. We are not related on any LLM which will be hallucinating every time. We'll be using this voice agent which is very good at its work and which knows his job and all. So we'll be justing we'll be relying on the voice agent and the purpose system will bridge the gap between the academic learning and industrial experiences. So this is the reason we this is the main thing we are tackling the gap between the academic learning and the industrial uh hiring expectations and all these system will uh will tackle the industrial uh hiring expectations and all. So coming to the problem statement let me let me also give you a concise thing where it will be u uh many students will be struggling with the uh prepare effect uh struggling to struggle to prepare the effectively for the interviews due to the lack of personalized mentorship and practical expression I have uh said it before right like we'll be having some personalized academies for this and all soademies what they do they'll be having some mock interviews and all in that mock interviews will be having uh some like bashed interviews and all mock interviews where user might not got the might not get the biased uh feedback and all. So we just uh we are just we have just developed a interview platform which will be coming with all the things that will be tackling these issues and all. So the manual interview requests human resources time and continuous availability. So this is the main big concern for the manual academics I mean academies which exists in the exist here in the societies. So what they need is the complex human resources they need for human resources and time and continuous evaluation and availability which will be not feasible for more than 10 users per day. So uh through our um uh web platform we can actually uh scale up to like 20 to I mean 10 thousands to millions of the users per day. So they can practice their interviews and um feel confident after their in their real interviews and all. So this is a problem statement that exists. Uh so they will be tackling up this problem in this uh platform through this platform.

[02:31:38]So and uh these are objective of the objective of our interview preparation web app application. So to develop an AI based intelligent system preparation platform. Yeah I've said it before that there this system will be uh intelligent uh intelligent uh interview preparation platform as it will be having some AI agents which will be aligning with the uh trustworthy AI and all. So we'll be uh providing the realistic voice based mock interview experience for the users.

[02:32:07]So we'll be using WP agent for that which is well known for its uh texttospech conversation and speech to text conversation which will be having some very fine tuned model for that. So we'll be not having any uh hallucinations in that it will be just doing its job very perfectly and precisely and to analyze the candidate responses using AI powered generation we'll be using LLM and rank powered LLM I mean rack powered LLMs which will be not hallucinating every time which will be giving you based unbiased evolutions and all and uh this will help you to improve your communication skills confidence and interview readiness. So what the students do is they'll um struggle a lot with their interviews and all. What they uh do is they'll uh start to mumble when the interview real uh realtime interview scenarios going on.

[02:32:57]Uh practicing on these mock interviews will be getting the user with the communication. We'll be getting some confidence for that and interview readiness will be guaranteed by this platform and to create a adaptive learning path based on the user strengths and weaknesses. So these were the objectives on the of this project and uh coming to the traditional reason we have focused on the gaps uh we have focused on the research gaps which were existed before the thing is the traditional uh which will be academics will be academies will be having uh as I said before this will be the same thing I have been repeating again and again they'll be having some uh issues with that thing concerns and all like uh they'll be having they should provide the continuous human interactions and human resources and uh they have to give some time I mean they have to give more time if if it's like uh 10 to 15 users I mean 10 to 15 students per day it will be very uh crucial for this for this to handle them so this voice system will improve the improve the scalability like thousands of users can actually use this and they can improve their communication confidence and all so uh this is what the literature research gap we have been working on uh methodology it requires analysis which consists the users needs and interview preparation. AI voice agent was integrated in this speech recognition to text to speech technologies. As I have said, we have used Wy AI agent which is good at its work and it will be it will be offering us some um low latency things and all. So, we'll be having some good uh benefits from this uh WP agent and this the reason we have used uh specifically WP. This is the reason we have uh choose for this. Everything we have chos uh everything we have chosen in this uh is precisely chosen for this thing not randomly we have chosen we have chosen precisely for every part of this project and uh coming to the system architecture model architecture uh as it is as it system architecture we'll be having file layers for this user front end back end voice and voice services and data storage coming to the user user will uh like user will be starting everything will be starting with the user user will login and Login will be handled by clerk. We'll be relying on clerk for this and also upon this we'll be relying on firebase fire store also for this thing. We'll be putting whole authentication process and uh authentication and authorization burden on the uh firebase thing fire so which is very uh we can rely on that actually we can trust for the trust their systems and all because it is very highly secure and all everything and it offers us offers with realtime database and all.

[02:35:36]So clerk is very trusted for uh this thing authentication and all. So we'll be using clerk which will be uh increasing our development speed and all and development speed and cost which will be affordable for users and all. So this uh the the flow starts with user user will be authenticated uh like based on his uh uh email and all he'll be having form validation and all everything and user after user uh is authenticated he'll be directed to the front end page which be having homepage.

[02:36:04]One page will be having some different cards. So user can select uh one of the multiple cards. The cards will be of technical non-technical HR behavior or interviews. User can select one of the card and he can start his interview preparation. What happens is after user selecting the card he will be directly I mean I mean back end is involved in this and the uh call will be API call will be directly hit to the WP agent and WP agent will come uh and come into the action here and WP agent will be uh taking care of the interview first phase of the interview will be it will be asking the questions like which technical thing you have to focus on how many number of questions you have to uh ask you and uh what is the uh level of the questions we have to ask you there will be few levels right? Freshers, entry level, senior level. So these are the things that will be connecting I mean collecting the data of this thing and it will be sent to the uh Gemini.

[02:36:58]Gemini is the LM we are using here. So it will be just doing the text to speech translation, speech to text translation, TTS and SST uh ST. Uh so we'll be using this for uh this thing. And uh here we have WPI for voice agent. So WPI is generally a voice agent. voice zoom where uh it will be uh giving us some benefits like low latency and all low latency and realtime uh human conversations and all. So it will these are the benefits we are will be getting from the WPI and it will be affordable for users if it's being published like it will be deployed and all. So after that we'll be authenticating with the data storage and all. We'll be having some uh this thing uh fire store for authentication and we'll be having firebase Google firebase which have which provides us with a realtime database and uh realtime authentication and all and firebase storage and all. So this is the thing we have here uh storage and user flow will goes like user will login using clerk and uh he starts the mock interview. The voice agent will uh uh voice agent will be handling everything like it will be asking us how many questions you have to answer. I mean how many questions I have to ask and how many how what is the level of the question and technical difficulty and everything will be asked and it will be collected by the agent and it will be sent to the Gemini and it will be posing that using is a well-known LLM and it will be sending to the W again and the next phase of the interview will be starting in the next step of the next stage of the interview what it does is it will be start uh it it starts the interviews and all so in interviews you'll be asking the based on the user collected objects and It will it will store the tokens and all everything for that objects and all. And uh after interviews interview is done uh the evolution engine evolution engine and feedback engine will get into the action and it will be giving it will be giving the real-time evolution and feedback. So where user get unbiased evolution and feedback based on what have the evolution will be on uh based on this um confidence uh technical knowledge u technical knowledge and everything and this is the system architecture how the user data flows and all and we'll be having results and discussion uh discussion uh section here. So here it will the feedback of feedback on the interview uh feedback section where after the interview is done we will be having some feedback where it will break down it will be the complete breakdown of the evaluation of the user user confidence and all and the technical knowledge depth and all. So you'll be uh he'll be having some uh score like based on on the scale of 100 you will be scoring everything and it will be checking every particular this thing and uh this will be the interview uh scenario where it will be asking everything question and this is the WP agent interview and this will be the person who will be getting interviewed and the system will successfully conduct the realistic air mock interviews. User will instantly receives the feedback after his interview is done. So uh this is the system and the conclusion and future work essay demonstrate the practical applications of the interview and the voice agent voice based mock interviews improves realism and user management and it will tackle the interview scenarios where it will be building the students with the confidence and all. So user can confidently attend his interviews after using this mock interview sessions and all. So it will be also reducing the cost of you uh cost of the students who can't afford uh the real mock interviews from the academies and all. So uh this is the reason we have worked on this project and uh uh future enhancements will be uh like we'll be scaling up to the m multiple I mean more number of the users if this get works and uh integration with real world hiring platforms and resume analysis system we can enhance his revenences and we can uh future work we will focus on the real time resumeum analysis and all so these are this thing and uh references we have referred these papers for uh for which we thought with these papers will be helpful for our uh project and we have referred these papers and uh we have uh build this project and this Thank you.

[02:41:10]So you're using open gemini means what is your contribution?

[02:41:14]>> So no sir this is the LLM which will be giving us >> you are using gemini only. Yeah.

[02:41:20]>> Then what is your contribution?

[02:41:22]>> So it will be like this engine will >> you are making front end design and giving to Gemini. Actually Gemini will take care of everything.

[02:41:31]>> Yes sir. Actually >> then what is your contribution >> sir? Actually this is not about the technical lab. It is actually about the about helping the students to build their confidence and all. We'll be using the API to itself. This is just basic project. These are many project to be honest. uh to be very honest this a very basic level of project and uh this is just the API toys just we have built it uh to uh for the students to help uh which will be we thought it will be helpful for the students to build their confidence and everything for this project some small drawbacks might be there >> yes sir >> sometimes a will do misjudgment of this candidate >> sorry sir >> A can do misjudgment of the candidate because I be good actually when I'm going to a model a model some other results it sometimes the a may misjudge about me >> sometimes there is no ethics >> yes sir >> you handle such kind of >> because when I do interview you are presenting I'm listening I'm seeing your eye contact how you're presenting where you are struck everything I observe in person >> yes sir >> if I'm a a >> Yes sir Sometimes I'm a misjudgment of the uh my answers. Sometimes I have some bias.

[02:42:52]Sometimes I don't have ethics.

[02:42:54]>> Yes sir. Actually I've said right sir this will be having this this system has drawbacks like it is a mini project which is a very basic project.

[02:43:03]>> Okay.

[02:43:03]>> Yeah. Uh that's the reason we have.

[02:43:05]Thanks.

[02:43:30]Please try.

[02:43:45]Don't for Hello.

[02:44:05]Uh my paper title is crisper argumented biofabrication with AI optimized genomic uh sleing and artificial room technology. This is the new technology actually a proof of concept development.

[02:44:20]We are extensively working on uh literature review how exactly techn technology is developing for the future like uh we are all suffering with so many disorders maybe covid-19 heart disorders neurological disorders manyophenia but so many mental disorders physiological disorders are somewhat common across the globe. So this particular uh concept proof of concept project is about how best we can make artificial womb which is uh developing a very sophisticated uh womb technology to prevent probable disorders for the human being.

[02:45:13]Yeah, my presentation uh first introduced what this topic all about.

[02:45:18]Then uh extensive literature we have done. Accordingly, we designed a methodology for this particular project.

[02:45:25]To support the methodology, we developed a few mathematical models and finally we discussed some of the results and finally we conclude with future possibilities of this artificial womb technology.

[02:45:38]So this crisper is like adopted bacterial immune system with modified RNA gene editing uh to modify RNA uh to direct cast 9 enzyme to the specific uh uh development of uh the DNA combinations which induces double strand bl enabling the gene insertion, gene deletion, gene correction and possible modifications at gene level and it has already proven that so many uh systolic fibrosis, cle cell diseases, so many disorders are already cured with this particular technology in uh various uh uh animal models have done and even in agricultural models also we have experimented we have seen the literature and I will give you some of the case studies about this and which combines five cutting edge technologies in a unified framework. I will show this framework in the next slide which bridges AI, genomics, stem cells tuning, 3D bioprinting, artificial wounds to develop artificial wounds. We are developing we are using these technologies and it targets for next gener next generation regenerating medicines as well as synthetic organals.

[02:46:57]These are the five different dimensions for this particular technology development. So this is the framework of the uh proof of concept work where the sample of uh the organ animal plants all are studied as a sample type. Then with the influences of virus, bacteria and so many infections with the host uh there are the possibility of variations at the gene level. Then uh is it possible to support by means of medication? Is it possible to cure? That is one dimension.

[02:47:32]And other dimension is is it possible to modify the gene itself. In that in that context we developed these underlining uh uh ideologies like uh resequencing the genes with the help of K means technologies and suitable graph and then finally we develop the unsupervised models and supervised models so that uh the required gene sequences will be reorganized to correct the problems. So I think sorry this is the literature where uh how the different people are working in different dimensions for foundations biofabrications stem cell engineering bor printing AI related RNA design and artificial ultimately we are looking at artificial womb development how these dimensions are supporting to this uh final artificial womb development These are the five six different phases where we need to do the gene sequencing at the first stage. Then AI optimized DNA RNA design is required as a second phase and then crisper cell editing is required and this is applicable applicable for stem cell modifications also. Then tissue biofabrication is in the fourth cell, fourth phase. And finally the fifth stage where artificial wound development is done externally and then it is clinical trial and validation is required for the uh future uh problematic uh faces in the animal health or plant health.

[02:49:16]So these are all the uh how DNA design factors are influenced for development of artificial womb with the GC content, seed regions, RNA length, target scrolling and then palm site and AI platforms are used in a software domain and these are considered as hardware domain and DNA pathways are used where in two dimensions for homologous and homology directed repairs. These are the two different dimensions for uh RNA design. And uh this is the how exactly the damaged DNAs are identified and then which are repaired by means of either biological donors or artificially developed DNA are considered here and then it is remodified and it is infected to the system so that effective system is introduced to make the system very efficient.

[02:50:13]Okay, this is how the physical organ system is considered and which is artificially developed. Real organ is considered as a reference. Accordingly, the artificial uh organs are analyzed and developed with the support of bioactive tissue model along with the subjective analysis through doctor patient communication, introspective navigation, performing experiments, prospective planning, simulations, medical evaluation, device casting and training skills are used to do some cell reactions, cell therapy and uh situation detection, drug screening, local tissue filling. All these factors are considered for 3D natural bioprinting or artificial 3D bioprinting for medical field with the uh sandwich of engineering field. This is what the different dimensions of uh technologies are working together parallelly for precise control of pH and temperature for inducing this artificial DNA into the womb and nutrient oxygen requirement as very essential component for physiological and anopical development of any organisms. So realtime tracking development progress is monitored in this artificial womb technology. So this is what some results which are which are all collected from the available literature and these technologies are already there across the globe with uh injured bioprinting and holo halo fiber bioreactors stereo lithography.

[02:51:58]This garage mixers these are all some of the parallel technologies.

[02:52:35]Good afternoon everyone.

[02:52:37]I'm Sai Rahman Satan and I'm science of engineering and management science to present present our research paper artificial intelligence and machine learning by various application impacts challenges and future prospects introduction artificial intelligence and machine learnings are reagenting modern libraries is by transforming traditional library systems into smart inter knowledge centers. These technologies help libraries automate routine activities such as cataloging, classification and information travel. AI improve search accuracy and enhance user services through chat boots and virtual assistance.

[02:53:32]While MLA machine learning enables user behavior to provide a personalized recommendations and better access to information. As a result, libraries are becoming more efficient, userfriendly and technologydriven in the digital era.

[02:53:49]Brief concept of artificial intelligence and machine learning in libraries. In artificial intelligence refers to the ability of machines and computer systems to perform task that normally require human intelligence such as a learning, reasoning, problem solving and decision making in libraries. AI used for a smart search systems, chart boards and automated cataloging services. In machine learning, machine learning is a branch of AI that enables systems to learn automatically from data, improve their performance without expert programming. Machine learning. Machine learning helps libraries provide personalized recommendation and efficient information.

[02:54:35]Retrol services.

[02:54:38]Objectives of the study to examine the role of importance of artificial intelligence and machine learning in modern libraries to analyze the applications of artificial intelligence and machine learning in library services such as a cataloging classification information retal digital libraries and smart library management to study the impact of event AIdriven technologies on library professionals users and knowledge management systems to identify emerging trends in AI enabled libraries including chatboards, recommendation systems, predictive analysis and smart libraries to discuss the ethical, technical, financial infrastructural challenges in implementing AI and ML in libraries to explore future opportunities and prospective prospects of AI and ML developing intelligent and userentric library systems to Understand how AML are transferring traditional libraries into smart knowledge center.

[02:55:47]Literature of review in the artificial intelligence and machine learnings are rapidly transforming traditional libraries into a smart intelligent knowledge systems by automating cataloging, classification, indexing and information retal services thereby improving the speed, accuracy and efficiency of library operations. AI technologies like chatboards, natural language processing, recommendation system and virtual assistance enhance user experience by providing personalized past and 24 into seven library services.

[02:56:25]Studies by Russell Norwig, Michael and Manning provide a strong theoretical foundations of AML applications in libraries, especially in areas such as semantic search, intelligent and information, literal, text, mining, and predictive analytics for digital library systems. in the literature to also emphasize the growing role of AI research about services including plagarism detection, citation analysis, centrometics, digital preservation and automated metadata generation which significantly support academic research and institutional repositories.

[02:57:19]Applications of artificial intelligence machine learning libraries. First one intelligent cataloging and classifications. Smart information retrol systems. Chatboards and virtual reference services. Recommendation systems. Digital libraries and knowledge discoveries, research analytics and centrometrics, smart library management, OCR and digitization, plagarism detection and research integrity, assetive technologies, methodology, literature review and data collection, concept conceptual and analytical study, analysis and practical applications, explority, evaluation, trends and challenges.

[02:58:04]Scope of the study to examine how artificial intelligence and machine learning are transferring modern libraries and library information science services. AML technologies are changing traditional libraries into smart digital knowledge centers. improving library services, automation, search research support and user experience while also addressing the challenges involved in adopting techn these technologies.

[02:58:31]Emerging trends in AIdriven libraries, AI ethics frameworks, smart libraries and intelligent buildings, generative AI and research writing and reference services. AI provide discovery platforms and knowledge graphs. Predictive analytics for library decision making.

[02:58:46]Integration of AI with IoT and RFD technologies in libraries. Personalized learning and research support systems.

[02:58:54]AI ethics and governance frameworks in LIS. Autonomous library and smart shelves. AI AI based digital preservation systems. Human AI collaboration in information services.

[02:59:05]These are all emerging trends in AI artificial learning and machine learning.

[02:59:17]Findings improved efficiency better user satisfaction satisfaction faster services reduced workload improve the service quality enhance user experience datadriven decision making ethical and privacy issues digital device challenges in implementing AI and machine learning in Library science, lack of technical skills, infrastructure cost, alder algorithmic bias, corporate issues, resistance to change, privacy concerns, ethical issues, high implementation cost.

[02:59:53]Conclusion: Artificial intelligence and machine learning are realizing the library ecosystem by enabling intelligent, efficient and userentric services. These technologies support advanced knowledge discovery, improve operational workflows and enhance research capabilities. Despite challenges related to ethics, privacy and infrastructures and ML offer immense potential for the future of libraries institution that strategically adopt these technologies will evolve into smart knowledge hubs playing a crucial role in the digital information society.

[03:00:29]These are references used to our papers.

[03:00:35]Thank you.

[03:00:37]Okay. Any questions? Thank you, sir.

[03:00:41]Thank you. Next paper.

[03:00:44]T3005.

[03:00:53]Next. T3007.

[03:00:58]007 are there >> and next T129 >> objectives based on this framework our core objectives are straightforward. It is to reduce active power losses and to improve the voltage profile at each bus buses and exa extract the precise optimal sizing for both configuration and sh capacitor. But by implementing this on a standard benchmark system, we provide clear reliable data to guide ultimately planning and power quality enhancements.

[03:01:37]Our work builds upon a foundation of literature while researchers have extensively used metahuristic architectures like PSO, GA, JR and greywolf algorithms and multiple hybrid algorithms.

[03:01:53]A clear gap remains regarding how algorithm selection directly impacts convergence and operational consistency when shifting from network geometry to physical assessment placement. This paper resolves the exact gap methodology.

[03:02:11]Let's look at the operational methodology. As mentioned, we utilize a rigorous two-stage approach. In stage one, we optimize the physical switching topology to minimize line losses. In stage two, we target the weak nodes, deploying optimization to find exact optimal placement and sizing of capacitor to satisfy reactive power demand. crucially to handle steady state network calculations with maximum mathematical accuracy. So we utilize the Newton reps and load flow method rather than simplified radical approximation and the placement which use are the fixed capacitor placement.

[03:02:53]We validated this framework on standard I3 bus distribution system as shown in the single line bus diagram. The su the system operates at base voltage of 12.66 kilowatt with a base of 100 mega volt ampers. It consists of 33 buses and 37 total branches and five normally open tile line uh represented by switches 33 to 37 and the results. Now let us discuss the most significant findings of our study which are summarized in table one and table two. The unoptimized base case starts with uh with a active power loss of 202.68 kilow and a voltage voltage profile of 0.913 PU. After reconfiguration of PS4, the active power active loss in stage one decreased to 139.98% uh kilowatt with a loss deduction of 30.935%.

[03:04:04]And the voltage profile of a bus I mean minimum voltage profile is 0.94129.

[03:04:11]After these stage two, the active loss of a PS4 again decreased to 95.773%age and a voltage profile increased to 0.96 pu. Looking at table two for geometric algorithm although geomet GA g performs well reducing ultimate uh ultimate losses to 99.6 kilow and h and achieving to 0.833% reduction. It is clear that PSO is con consistently outpaces GA in searching the complex solution space.

[03:04:49]In conclusion, our paper, our study successfully proves that while both algorithms yield massive ex uh efficiency improvements, PA swap stands out as superior more robust uh optimization engine for multi-stage diffusion problems. Moving forward, our future work will look to expand this unified framework to accom accommodate dist uh distribution generation placement. We aim to optimize the concurrent location and sizing of renewable DG units. Validate system scaling on larger networks like IW66 bus systems and introduced time varying dynamic load profiles to mirror realistic daily banks.

[03:05:36]And these are the fundamental differences that that we took. And thank you.

[03:05:44]>> Okay. Vul thank you for your presentation.

[03:05:47]>> Okay sir.

[03:05:48]>> Okay.

[03:05:52]Okay.

[03:06:06]May I speak to Kirana, Bangladesh?

[03:06:11]>> Uh yes sir, I'll share my presentation.

[03:06:27]Madam, can you call me after 30 minutes?

[03:06:40]Can I speak to Kana Wangadesh?

[03:06:43]>> Yes sir.

[03:06:46]>> Okay. What is your paper ID?

[03:06:48]>> Uh I the IC EA 2026_24 repeat it. IC A S A >> last >> uh 054 >> 054.

[03:07:14]>> Yes.

[03:07:15]>> Okay. Continue with your presentation.

[03:07:19]Myself Kana I'm going to present about the project that we have done that is the health care system for the paralysis patients and uh the main authors of this uh paper are our guide Mrs. Shubabi and our teammates Kithana, Kiti, Purvi and Bhavy Shri and uh we are from JS's Academy of Technical Education Bangalore. Now I'm going through the presentation outline that is what what I am presenting today. Uh I'll be sharing the introduction or the background behind our project and the problem statement that we have gained from this uh literature survey and the objectives of this project and the methodology.

[03:08:01]uh then the results and our and I'll conclude with the uh future scope of this project and the other uh results and what we are going to do next >> like integration with the 60 communication networks >> the impact of the paralysis on the individuals are very crucial and severe because they lose their mobility mainly they lose their independence and be there are many causes of the paralysis such as the spinal cord damage or some stroke or some other causes that is maybe due to the accident. The people will mainly lose their ability of many parts such as leg or some brain or the hand and many paralyzed individuals they are suffered the main difficulty they face is they left out with the adequate assistant from the caretakers. So in order to um avoid this problem or to give to address this issue we have come up with a project that is to develop an integrated health care system for the paris patient and uh this is the problem statement. As we know they are suffer they face common difficulty that is they are lack of adequate care from the caregivers. We have developed a centralized or an integrated healthcare system for them that is that contains ECG sensors, EMG sensors and the flex sensors in order to give or monitor their vital health parameters such as heartbeats or the muscle movement in order to give them a good rehabilitation and flex sensors in order to increase their independence. And uh the main objectives of this project are to develop an healthcare system comprising of various sensor and also to give a to implement a voice-based assistance to the caretakers.

[03:09:56]Then we'll go through the literature sub survey. The first paper the first paper that is IoT based robotic system for rehabilitation of paralysis patients. We have taken it from the journal international journal of advanced robotic systems. The main summary of this paper is robotic devices are connected to various IoT devices that can support rehabilitation exercises and it also monitors the patient progress remotely and helps restore most motor function in the paralysis patient. The main drawback of this paper was it was of a high cost of including the robotic devices and also integrated with the IoT. And the second paper was BCI that is brain controlled emergency calling system for the paralyzed individuals and it was taken from the IT E journal and uh the main uh summary of this paper was the brain signals and I go to the implementation.

[03:10:57]>> Okay sir.

[03:10:59]This is a block diagram of our all the main implementation of our um project that is it contained a processor that is the or aino uno and many inputs from the sensors such as flex LM35 ECG and EMG.

[03:11:12]The output we have got through the voice module or this and the speaker and the LCD. This is the flowchart of our proposed system that is the sensors will read the input from ECG, EMG and temperature and text sensors. LCG sensors are used to uh monitor the heartbeats and also the the output we have got through the serial monitor and are used to uh what is that the based on the finger you'll get the outputs such as I need food I need water I need medicine and all through the voice module and the ECG sensor is used to calculate the as they are suffering from as they may uh go through vital uh uh health then um the after all this uh uh re it will return to the system implementation. This is the proposed healthare system.

[03:12:15]These are all we have integrated together all these and uh this is the outputs of the ECG graph that is uh the heart rate is uh what is it? It is showed through the serial monitor and these are the BPM that we have calculated whether the heart is normal or it is high heartbeat or it is very low. We can show it through the serial monitor and this is the EMG sensor. It basically measures the muzzle activity of the paralysis patients. We have we will set a range to it of if the muzzle activity is more than that the sero motor will run. Based on this computation results, we can give rehabilitation to the paralysis patients. And third one is the LM35 sensor that is mainly used to measure the temperature and temperature of the patient and also the output or the temperature that can be apart from the Hello.

[03:13:15]>> Yes sir.

[03:13:16]>> Hello.

[03:13:18]>> Yes sir. Apart from these sensors, what is your contribution for this title?

[03:13:26]>> Sir, we have integrated all together that we that can can't be seen in majority of the papers that we have read.

[03:13:36]>> How you are able to develop that interface?

[03:13:39]>> Through ordino uno, we have connected all these sensors to a single controller that is ao uno. that uh outputs will be seen through the voice module and the LCD.

[03:13:55]Okay.

[03:13:57]Okay. Thank you for your presentation.

[03:14:05]You can stop now.

[03:14:23]May I speak to Sneha?

[03:14:26]>> Am I audible, sir?

[03:14:31]>> Hello.

[03:14:33]>> Hello.

[03:14:34]>> I'm audible, sir.

[03:14:38]>> Yes, sir. My paper ID can continue with your presentation. May I know your paper ID? 58.

[03:14:43]>> May I know your paper ID?

[03:14:45]>> Sir, my paper ID is 58. Hello.

[03:14:48]>> Ah, you may know your paper ID.

[03:14:52]>> 58.

[03:14:59]>> Hello. May I know your paper ID?

[03:15:02]>> Sir, my paper ID is the last three digits.

[03:15:05]>> 58. 058.

[03:15:07]>> 508.

[03:15:10]>> 58. Yeah.

[03:15:15]>> Can I start?

[03:15:15]>> Oh, okay. You continue with your presentation.

[03:15:18]>> Okay. So, my paper title is real time monitoring and intrusion detection in the smart system. So, the uh my paper ID is 058. Author's name are myself and my teammates Tanushi and Deep.

[03:15:32]>> Okay. Continue with your presentation.

[03:15:35]>> Hello. Am I audible, sir?

[03:15:36]>> Oh, okay. Continue. Yeah.

[03:15:38]>> Continue with presentation.

[03:15:40]1 minute.

[03:15:47]>> Okay, no problem. You continue with your presentation.

[03:15:51]>> Okay, sir. One minute.

[03:15:55]Uh so uh uh uh introduction see in today's interconnected word so the security tool cubical has important today every camera surveillance or anything has to have human monitoring which has time which takes lot of time and >> your object >> go to your objectives and implementation part. Okay. So my our objective of the paper is to develop a realtime surveillance system, detect and identify introduce intruders and instantane alerts and alarms and to capture and store evidence and remote monitoring and control. So how we implemented this is that so we used a motion sensor and we used a uh touch sensor also and a face recognition model and introduc intrusion detection and alert generation and data processing and the hardware and the software part. We integrated the hardware and the software so that the efficiency of intrusion detection will be more. So the block diagram is the >> so it will be like a motion sensor will be monitoring 24 hours. So what happens once a motion is detected the face will be recognized in the camera. If it is a the AI technology we would have trained the AI that is a familiar face everything. So if the face is not recognized what happens and if the behavior is not normal like for example if we take the world so the evidence will be taken and immediately the system will live hello yeah so immediately what happens the app will open in their phone or connected apps and immediately it will start live recording of the the screen or the system surroundings so that it will capture evidence and if the example uh if the system is in open and they are not there and some data is lost easily they can have evidence who has come and who has who are around the surrounding.

[03:18:05]So this creates a better safety for the smart systems.

[03:18:10]So the results uh face recognition accuracy was 94.7let notification was around 1.8 8 second immediately we got this we tested using the analyzed phase and not uh high intruder and then continuous monitoring was 24 hours so the conclusion I would say is uh it will the system what we build it integrates motion detection face recognition and realtime alert mechanism so the future scope is that this can be implemented in all the smart systems like camera or the laptops or the ATMs wherever ever using nanotechnology. So whenever the system is built, this can be immediately uh built in so that no need of human.

[03:18:58]>> Yeah sir. Hello.

[03:19:00]>> Okay. Uh tell me what are the uh parameters taken for face recognization.

[03:19:06]So for uh parameters that we have taken is we used the uh AI model uh that is we took the like around 500 images of the same person and also 2 three four images of different person and train the AI model.

[03:19:25]>> From the images what are the parameters?

[03:19:28]Parameters means what are the properties you are going to take for the consideration? uh properties are the facial features.

[03:19:38]>> Yes, facial features and uh the exact feature or the eyes everything.

[03:19:50]>> Okay. Thank you for your presentation.

[03:20:06]May I speak to Shamba?

[03:20:09]>> Sha.

[03:20:10]>> Yes sir.

[03:20:12]>> Uh may I know your paper ID?

[03:20:14]>> 052 sir.

[03:20:17]>> 052.

[03:20:23]>> Okay. You continue with your presentation.

[03:20:26]So directly go to the objectives and implementation part.

[03:20:30]>> Yes sir.

[03:20:35]>> Sir my is my screen visible.

[03:20:44]>> Excuse me sir.

[03:20:45]>> Yeah it is visible. You can continue.

[03:20:48]>> Yes sir.

[03:20:53]>> Uh good afternoon sir. I am Shama from Jim University first year BTech. Uh uh so today I'm presenting uh my paper of design uh design and implementation of 4bit flash ADC flash ADC using uh 90mm technology. So my objectives uh my main objectives are to design uh to design power efficient 4-bit ADC using 19M CMOS technology and to reduce the power consumption by replacing R2R ladder with the sample and whole circuit to simplify and comparator and encoder design to improve the speed and efficiency to implement and simulate the design using cadence purchaser. So next we will go through meth methodology. So the prop proposed methodology focus on designing a low power flash ADC by optimizing the major circuit blocks components. Uh we the four main components are sample and hold circuit compar discrete comparator priority encoder and test circuit.

[03:22:08]Next we will see the our circuit like sample and hold circuit. So samp sample and hold circuit is first stage of the ADC. It function is to sample the analog input signal hold signal and constant during conversion.

[03:22:28]This improves the stability and conversion accuracy in our design.

[03:22:33]replacing the R2R ladder with the sample and circuit help to uh reduce the circuit complexity and power consumption. So next we will go through the discrete comparator. This is the schematic uh representation of a discrete comparator and this is our output waveform. The comparator compares the sampled input wtage with the reference wtage. If the input wtage is greater than the reference wtage, the output becomes high. otherwise uh it remains low. The comparator plays an important role in high-speed conversion.

[03:23:10]So in our project uh the simple comparator structure is used. It provides a faster switching. It reduces the uh delay and power consumption. So next we designed the priority encoder.

[03:23:26]Priority encoder is a digital circuit that uh converts the multiple input signal into binary output code. If more than one input is active active, the encoder gives the priority to the highest order input. For example, if both two and three are high, the it is in higher priority.

[03:23:54]So advantages of these priority encoders are reduce the number of number of outlines, improve the efficiency, avoid the output uh ablity. Priority encoders are commonly used in microprocessors and uh communication systems. So next uh at last we designed the test circuits. Uh this uh this complete flash ADC circuit was tested using CAD virtuoso. The output waveform uh confirmed successfully analog to digital conversion compared to traditional 180nm technology. R90NM is showed low power consumption and reduce the propagation delay, faster switching speed, better integration uh density. Thus the proposed ADC achieved improved overall performance. So uh this is our comparison table. So uh this is our work and this is the compare uh we compared this 90 mm technology with 180 mm. So frequency we took is 1 kilohz technology is 90 mm is 1.8 resolution is R4 bit delay is 6.8 which is half of the 180 mm technology and power dissipation of sample and whole circuit is 0.040 040 mill which is also half of the 180 and power dissipation of comparator is 0.036 matt which is also half of the 180 nm technology power dissipation is of encoder is 6.58 into 10 ^ minus 8 so by this I'd like to conclude we successfully designed a power Hello everyone. Good afternoon. Myself Sani from Global Academy of Technology.

[03:26:11]Uh uh my paper title is uh self balancing two wheel robot for medical logistics. uh we prepared the paper and project under the guidance of uh professor assistant professor Shouba Gian and uh we are a group of four members.

[03:26:28]So so the the introduction is about >> sorry uh so these are the presentation outline that I'm going to present. So the introduction is uh like in the modern days uh uh robotic systems have gained a significance importance in industrial areas and in uh even in the healthcare facilities.

[03:26:58]uh uh due to their ability to improve efficiency, reduce manual effort, uh we are developing this uh prototype. uh uh we are mainly uh developing this in a hospital environment as um there will be some uh uh infectious uh as it is infectious environment uh it would be helpful uh for the humans uh and uh it would also give the speed uh manual manual variability.

[03:27:29]self- balancing two wheel robots uh have emerged as an effective solution for indoor logistics uh and also ability uh it has ability to navigate through narrow spaces. So the mainly the problem statement is uh hospitals and healthcare environments face delays uh inefficiency inefficiencies and um also risks in medical supply transport. Uh so in the as the hospital environment is very uh compact and um there will be narrow surfaces this uh prototype self this robot can move in all the uh all the desired environment. So the objectives are uh uh to develop a cost- effective two-build uh robot with P control uh and uh indoor navigation where P uh control is the proportional integral derivative uh which uh which which means uh which uh which means uh it uh uh sorry um it eliminates steady state errors and also uh it give uh it measures uh uh deep tilt angle of the robot.

[03:28:38]uh to integrate wireless uh communication and a user friendly uh interface for remote monitoring and also to include safety features like bazar alerts uh and uh like uh implementing ultrasonic sensors for object detection.

[03:28:53]So the literature review as uh the methodology uh the proposed system the proposed self balancing robot uh is a mechanism based on inverted pendulum where it needs to balance only on two wheels. Uh the MPU 650 sensor uh we have used it continuously measures the robot's tilt angle and angular velocity. uh the P controller uh where I told uh proportional integral derivative it calculates the error between the desired and actual tilt angle and helps to balance the robot. The calculated output is sent to the DC motors uh where L298N motors are used here and uh the ultrasonic sensor is used for ob obstacle detection. Uh so the complete system operates in a closed loop feedback structure.

[03:29:45]uh the system architecture is about uh ultrasonic sensor where I told uh uh which detects the objects uh uh uh which come closer and uh move backward the robot move back moves back backward. IMU is the initial measurement unit and microcontroller is the IC main IC uh uh P controller uh uh the main IC uh is send sent to the P controller which controls the tilt angle of the robot uh and uh this is uh uh as we have also used Wi-Fi communication uh it helps us to navigate to the hospital uh environments and these are given to the motor drivers and gear motors.

[03:30:30]So the results and discussion are uh this is the developed self-balancing robot we have built. Uh the P controller has achieved stable self balancing and smooth robot movement. MPU 650 uh has provided accurate real-time tilt measurement for balancing. Ultrasonic sensors have enabled effective obstacle detection and collision avoidance. The robot maintain balance during forward and backward motion with minor obstacles and oscillations.

[03:30:59]So the references are uh these are the references we have u uh chosen from and uh thank you.

[03:31:08]>> Why only two wheel? Why you have go for not four wheels?

[03:31:11]>> Uh as I told sir in the hospital there might be narrow conditions or even uh uh the big prototype cannot move as four wheels is a big prototype uh it cannot move in that uh places. So >> anyhow you are carrying medicines and all right >> it should balance now >> sir it will balance it will balance >> two two wheel robot will balance >> what is the weight you can carry >> uh now as it is a prototype uh it will carry about 1 kg but we are uh as it is a phase one uh we are still going to build it uh large >> okay thank you thank you sir next anyone from other colleges Heat. Heat.

[03:32:17]Heat. Heat.

[03:32:32]Heat. Heat.

[03:32:52]Next material pretty Yeah, even Uh, Shahini Sharma.

[03:34:29]No.

[03:34:46]Okay.

[03:34:48]Wonderful.

[03:35:16]So good afternoon everyone. I'm Shahini Sharma from um second year AML department at Ambutra Institute of Engineering and Management Science and my research is based on aggression machine learning based ransomware attack detection in healthcare system. So this will be the topics which will be covered in today's sessions. So nowadays everything is getting digitalized from banking to hospitality and mainly healthcare is mostly dependent on digital data. So digital data has increased the risk of cyber uh threats and especially ransomware attacks in healthcare systems. So ML techniques, ML techniques and machine learning can be used to solve this problems.

[03:36:03]So in problem statements uh let's go with an example. Suppose a patient is having a issue and he went to he went to uh hospital to check the issue and he he is diagnosed with uh cancer and but some attacker have attacked the system and changed the data and so in this case risk will be more.

[03:36:33]So objectives of this research is to study the impact of ransomware data in the healthcare system to develop the machine learning techniques and check whether it is working on the well for healthcare industry or not. We can uh compare the difference between unsupervised and supervised learning and we can eval evaluate the graph metrics based on these other uh we can evaluate the model performance based on accuracy, precision, recall value and etc. So in methodology we collected a data set from Kaggle realtime data set. We uh processed it. We uh removed the unwanted data sets and uh duplicate data sets.

[03:37:21]Then we trained them using supervised learning and unsupervised learning. In supervised learning we use logistic regression, random forest, artificial neural network and support vector machine. So supervised learning are used for uh so system architecture. This is the first we collected the date.

[03:37:44]So in result uh out of four supervised learning random forest and ENLN have achieved a very good results and in unsupervised learning came in crusting and hierarchal crusting were very well in detecting the abnormal activities which has been uh detected in the systems and DB scan and mean shift were they performed poorly because they were not able to differentiate between the differentiate between the plaster that whether it is uh abnormal or normal activity.

[03:38:21]>> Do you think needs to be used to work on this doain?

[03:38:26]>> Yeah.

[03:38:28]>> Okay. Thank you.

[03:38:32]Next.

[03:38:52]Heat. Heat.

[03:39:23]Good afternoon everyone. This is Prrii from AML second year studying at Amita Institute of Engineering and Management Science. So my paper title a holistic framework for rooting security energy efficiency and performance optimization in mobile adopt networks. So these are the presentation outline which includes with slight sense.

[03:39:54]>> No, not like that.

[03:39:57]>> Okay. So, what do you mean by manet? So, as soon as we hear about manet, one thing will get in our mind is uh uh devices is acting as both host as well as the intermediate node. So, let me explain with an example. Imagine you are somewhere in city or somewhere and you lost your device. So here how we going to find the device that comes our first question. So using magnet we can solve those problems not only in personal things but we can also implement it in military uh disasters or somewhere in >> Yeah. uh these are actually these uh since this is my first paper publishing uh I have taken these videos from uh papers which have already been published so I'll explain this with the architecture actually yeah uh so here comes the architecture green as which is it is a it's a thing which calculates the path every path in a sense what and all parts we have what and all options we have it will it will select all the parts It will find identify every part present over it. And then comes the T which which calculates the cost in a sense the it calculates the cost where we find you can see here.

[03:41:29]Yeah.

[03:41:31]Uh you can calculate the cost weight.

[03:41:33]You can see the how much time it is taking or what is the percentage of time it is taking when it comes to the AMS it is related to the energy conservation where it has been proved that 42% after using APS it is proved that 42% of energy conservation we can do and it has been successful also since it is published in other papers also it is it has been confirmed then comes to CLM where It holds the complete data set. If example, if we are if you are in a disaster area or some flood have been occurred and we want to do a rescue session.

[03:42:19]>> Okay. Okay. Um uh if we are in a disaster situation and we are going to rescue a person. So think there is a multiple robots or someone and they need to find the that person particularly and they need to find every path in a sense they have to search everywhere the path wherever it is present. So it can be done like one robot goes on one path and it will select the path and it will uh store the information whether that particular path is proper and it does not have any obstacle or something and then uh and the data stored by that one particular robot is shared with the other robots which are on the same mission. So in this manner it is uh man it can be applicable. Then comes the yeah this is the system architecture where you can see the source node you can see the source node and that are interacting with the intermediate node intermediate nodes.

[03:43:13]There are uh yeah they are interacting with the intermediate node and finally reaching the destination.

[03:43:20]Yeah, these are the observation got from the uh literature review.

[03:43:26]Then come the conclusion. So hence we can conclude that uh now mandate is not only used for personal gain or somewhat but it can be used in militaries IoT devices and etc. Yeah future work uh when it comes to future work we can replace social sectors with DQ and people and agent transfer place. So we can implement AI in this. We can incorporate AI and make it even more better.

[03:43:51]These are the references.

[03:43:54]Thank you.

[03:43:57]Questions.

[03:44:03]Doing it for first.

[03:44:05]>> Yeah. Come to first slide.

[03:44:18]And this paper is completely your uh review paper or survey paper not implemented not in result paper just you have compared different papers on >> this is the first paper which includes all the four challenges facing. So we have previously every single is a single paper but have combined all the four challenges and made it uh reviewed it into one paper. But what is the meaning of optimization?

[03:44:44]>> Yeah, it's like in a sense we are reducing the complex into simple and making it even more better.

[03:44:51]>> Okay. What are the applications of ad hop?

[03:44:55]>> Uh no I said no sir. Uh example you are in a flood. You are in a place where flood has been occurred and there is there are people who are missing and and you can take a military as an example and if the soldiers are far away from each other and they want to communicate with each other and uh and of course they'll be having walkie-talkies and if the third person who is far away from the first person and in a sense a walkie-talkie can reach only till the 100 m radius of it. So if you want to if that particular person want to reach and communicate with the third person he can uh communicate with the middle person and then reach the third person but one advantage is in here is the middle person who is present they won't be knowing the any messages uh because in uh middle person their IP address in the bucket okay whatever the information it is if it belongs to them then only going to check or else it going to pass the signal so it going to reach from source to destination properly without any u malware or malicious things.

[03:45:56]>> Okay, thank you. Next Still how many are pending to present?

[03:46:31]All presentations are over.

[03:46:36]Okay.

[03:46:46]CS paper 17 T27 Good afternoon everyone. I'm Niserga from the department of artificial intelligence and machine learning and my research title is uh deep learning based identification of uh microplastic contaminants in the organic waste using a match processing technique. Uh this focuses on the segregation of microplastic from the or the sura.

[03:47:35]This focuses on detecting the microplastic in the organic waste.

[03:48:07]Uh in this presentation we will discuss about all of this. Uh introduction microlastics are the plastic particles which are lesser than 5 mm that are found in compost and organic waste, soil and everywhere. These are the major problem which we'll see here. The there were traditional methods which were there to segregate the microplastic from the organic waste. But they are they required more time and they are uh high cost.

[03:48:43]Then problem statement microlastic that are present in organic waste affect the compost and ecosystem environmental sustainability. Uh there are some existing methods that were there to detect the microlastic but they have some limitation. They require the expensive equipment to detect and uh they require long processing time. Then we have to trying uh more complex sample.

[03:49:16]Then high dependence on dryer operators.

[03:49:21]Objective. The main objective of our uh research paper was to identify the microplastic present in organic waste and to apply image processing technique used to deduct the microplastic and to develop the CNN model for the uh classification of contaminated and non-contaminated uh Yes.

[03:49:52]Non-contaminated waste detection literature view.

[03:49:58]Next methodology methodology. We have some several steps here for uh how our system will work.

[03:50:10]First we will collect the organic waste images and we will uh pre-process that image. After pre-process we will send that to the CN model. The CN model will uh extract the future work like a texture shape. Then then it will uh contaminate and it will produce the result it is contaminated or non-contaminated.

[03:50:35]Then this is the system architecture.

[03:50:42]Then conclusion and future work.

[03:50:45]Microplastics are the now currently become major several challenge. This research proposed the CN based model for the detection of microlastics which is helpful in u uh detection of microplastic.

[03:51:02]It will consume less time and it is low cost.

[03:51:08]future work in future we can uh do real models using long data set realtime detection systems and improved CNN architecture then these are all the references which we took and this is a review paper like u the less researched area was this so it is research >> okay my question is how exactly you use image processing technique in detecting microplastic >> because how it is visible in images.

[03:51:38]>> We took images like uh >> how it is visible. Microplastic means what? What is the meaning of microplastic?

[03:51:46]>> Uh the uh cut down of plastic large >> what is the size of that?

[03:51:50]>> Uh 5 mm.

[03:51:52]>> 5 mm. How it is visible in image processing image? Uh what uh what you take is that visible?

[03:52:00]>> Yes. Uh in CN model it will visible.

[03:52:04]>> Thank you. Anyone else?

[03:52:07]Okay, thank you all for your uh patience support uh for conducting this uh conference and al I congratulate all your uh the crew members for this. Thank you all and also our uh online uh uh sir who is uh very busy for 2 days completely engaged in uh what uh streaming our videos. Thank you sir.

[03:52:39]Give me a minute.

[03:53:07]Start again. Start again.

[03:54:28]made it.

[03:54:30]Hello. Hello.

[03:54:33]Please come on the stage.

[03:54:38]I would like to request Baki from Hello.

[03:55:19]Hey.

[03:55:34]Hello.

[03:55:46]Vil 90 on 300 30 is equal to 90 + 230. That's it.

[03:56:00]>> Yeah, I know.

[04:41:10]Hello. Hello. Hello. Hello.

[04:41:13]Hello. Hello. Hello.

[04:41:17]1 2 3. No, it is not.

[04:41:28]1 2 3 check.

[04:41:30]Wake up.

[04:41:32]Wake up. Yes. Wake up.

[04:41:41]Fine. Perfect.

[04:41:44]Yeah. Thank you as well. Did you have your lunch? No. I had >> You're dieting from the last two days.

[04:41:50]It is like naratri for you. You have not had food from the last two days.

[04:41:54]Today you had relaxed perfect.

[04:42:10]So before the session starts katanori let me announce the winners of today's paper best paper.

[04:42:41]Relax, relax, relax.

[04:42:51]done.

[04:42:55]H >> yes.

[04:43:02]>> Yeah. Good afternoon everyone. Uh let me announce the best paper for the second day of ICT.

[04:43:09]So to begin with the first best paper of uh second day will go to T4044 design and simulation of RA RA for smart wireless technology and 6G. The second one goes to T1002 foot flow MIDI intelligent food prediction and the next one is T1056 towards trustworththing AI in diagnostic center and biometric uh landscape. T2075 uh it's an adaptive behavioral framework inside the uh threat prediction and uh the next paper is T2018 smart college app and the next one is T4062 autonomous wireless security the next one is T134 and a driven multisource and the last one is T401 and I request all the participants of the paper to wait until miss the validtory. The validtory will be deciding in another 15 minutes. So, hope to see you soon. Thank you.

[04:44:31]Announcement for the faculties as well.

[04:44:33]So it is better that you sit along with your committee members. If at all you belong to promotional team, sit along with the promotional committee. If at all you belong to uh editorial board that is uh whoever got the papers and reviewed, you sit with your team. If at all you into stage so let me just read out these names that is much more complicated.

[04:45:06]You can hear from both the mics, right?

[04:45:09]Huh? You can hear. Okay. So, sulpa coming. So, sit with your uh uh committees. So, to begin with, we have uh if at all you have your friends, please connect them and sit together.

[04:45:22]The publicity and media committee uh it begins with professor sanit uh Dr. uh Nagalandanda, Shinwas, Vinod, Rohit, Vive, Rohiteshwar, Nagata and Prasad kindly sit together kindly sit together editorial and technical program committee and uh uh Dr. Pratarvi, Professor Ja Prasad, Mangjanat, Cavia, Sangeita, Akata, Vinod, Nam Shi, Kavia and Rohit Kumar. So if at all you are into two groups it's you have to manage. I cannot say anything about it. The conference proceeding and publication committee Anita Madam Sanjana Kusama Vijay Jayashri was Arita Rakita Rohit Teshwar Nagata and Rohitkumar on stage committee Rupi Kithana Samya Prasad Kusma and Rashmi and Rikay registration committee Dr. Nagan Professor Prasad Prasad Prasad Prasad.

[04:46:32]How can you be in so many committees?

[04:46:35]So registration committee Naganandanda, Prasad, Braxita, Nam Shi, Reika, Rahut and Vive madam proceedings.

[04:46:48]Proceedings you have to tell me I was proceeding Anita Sanjana Kusma Vijay uh Shri wasant Arpita Rakita Rohit Nagapa and Rohit Rohit Kumar you just add your name as well.

[04:47:15]Have you and with respect to the stage committee, it's Let's go and with respect to the stage committee we have Rupi Kana Samya Prasad Kusuma and Rashmi as well. So the registration committee uh Dr. Naganandanda Prasad Prasad again you are here Rakita Nam Shi Reika Rohit and Vive Duh and uh the next one what we have is the hospitality committee uh the professor Kavia Kabya Mangjanat Mangjanat uh Rashmi Axhata Aishwara and Rashmi the food committee is uh Vir Patel and uh Shar uh certificate committee Mr. Many Mangjanat and Vasendradi technical session committee Axhata Vinod and student coordinators here out your names if at all anybody's name is missing please let me know And the student coordinators please add if at all missed your names. I I'll read out the names. If anybody's name is missing please add it out. Kanesh, Danosh, Sujan, Siddhar, Yanesh, Abhishek, Abhishek, Sat, Vashan, Hitesh, Gagan, Punit, Adipita, Danushk, Gautam, Jibim, Ram, Chundra, Prasad and Priyanka. Done.

[04:49:54]So this is the committee. Please make sure that you people sit together and thank you.

[04:50:24]And uh the last one is uh technical support committee uh Dr. Chaitan Raong, Professor Anil Chaitan, Rakkesh, Shinas, Varun, Shilpan, Deepi, please sit together.

[04:52:43]Heat.

[04:52:45]Heat.

[04:52:52]Heat.

[04:52:53]Heat.

[04:53:16]Ever since I was a chick and I was shy keep me by my side.

[04:53:29]Now that I'm put my sun on red, I put off the hit on 22.

[04:53:40]So put on a shirt, get on a feel like a stain. Get fit. I had to tell her that I was a nose.

[04:53:49]She singing my songs. She want to die.

[04:53:53]Trying to get hit.

[04:53:58]Smile on my face.

[04:54:01]Come and get chick. Ice on my neck. I roll her wrist. She in the sky. Push out the ch.

[04:54:08]Yeah. I spinning in just for these fun.

[04:54:15]I'm ringing all of my viewers. I feel like ever since I was a kid.

[04:54:24]If I was you, I would cut my exotic all over my body. Oh, she just swing my ever since I was a I should let her go. She want to be tight on her body.

[04:54:48]It don't matter what they say on time.

[04:54:54]City on fire when I'm coming home. Fill up the sky. I'll fill up the play one day. It's a hell of a show, but it's going to hurt cuz we did it first.

[04:55:10]Still like skate party boys on the feel like it's all three.

[04:55:16]Neptune dr. She hit me flipping a be.

[04:55:20]She want to the team. She fell in love with a dream. She fell in love with the scene.

[04:55:26]Oh yeah. Her man quiet not a th broke his heart. PTSD hold his chest. Let him breathe. Let it breathe. Ste and I got a priest. He got a cross. Get out of line to God. I should have pray for the laws of you.

[04:55:44]Ever since I was a kid, I believe you.

[04:55:48]If I was you, I would call my exot.

[04:55:56]She swing my ever since I was a G. Are you I should let her go. She want to be. Oh no. Got my old t on her body.

[04:56:12]You don't say timeless timeless. Heat.

[04:56:16]Heat.

[04:56:26]Heat. Heat. Heat.

[04:56:36]Heat.

[04:56:42]Heat. Heat.

[04:56:59]Oh, heat, heat.

[04:57:17]Heat.

[04:57:32]Heat.

[04:57:40]Natal.

[04:58:00]I am roll the down.

[04:58:58]Heat up here.

[04:59:35]Heat. Heat.

[05:00:11]I don't do that to Hey Miami Honey, honey.

[05:01:28]Good afternoon to all and all present here. We have now come to the concluding ceremony of the international conference on trustworthy AI for intelligent computing systems 2026.

[05:01:40]It's my privilege to welcome you all to the validator session. The last two days we have been a journey of ideas, innovation and collaboration. To begin this session, we want to invoke the divine blessings for this validator session. So I request Rashmi and Sajja from second CSC to present an invocation song.

[05:02:21]Sh.

[05:02:36]Sh. foreign.

[05:03:12]foreign.

[05:03:29]Thank you.

[05:03:40]Thank you Ashmi and Sja for that beautiful invocation. A conference ends but collaboration begin. To formally welcome our devotes to this session, I now invite Dr. Kusama KB to deliver the welcome address.

[05:03:57]Dr. Shrari whose dedicated effort, planning and coordination made this program successful. welcome you also.

[05:04:06]Mantes >> a warm welcome to all dignitaries, faculty members, dear participants present here today as we gather here to celebrate a successful completion of this conference. I hope today victory session will be me memorable and inspiring all. Thank you and once again warm welcome to everyone. Thank you.

[05:04:35]Thank you ma'am. Every successful event has a story. To present a brief report on the highlights and outcomes of the international conference on trustworthy AI for intelligent computing systems 2026. I invite Professor Shss Hod of AI and ML and basic science department to the podium.

[05:04:57]Principal Dr. Santo Shaman Murman, Dean Academics Dr. Raj dev cardinet my fellow conveners distinguished session chairs keynote speakers faculties and our presenters a very warm welcome to the val section of ICTS 2026 before the before the report we acknowledge two pillars behind this conference Dr. Virant the chairman being sa babul kotto and shri manti welcome address by rajikarmut release of conference proceedings facil facilitation of our chief guest and the inauguration concluded with a state anthem following the formal inauguration the keynote session is started Mr. Euron the chief technical officer NDI communication Israel delivered the first keynote on a driven intelligent computing architecture.

[05:06:03]He was followed by Mr. Nandi Vanakishor the founder and CEO of Akar Ash ashtaksha labs Bengaluru who spoke on secured intelligence system from the edge to cloud. Post launch two backtoback parallel session presentation sessions ran from 2 p.m. to 3 p.m. and 3:45 3:45 p.m. to 4:15 p.m. across five simultaneous tracks in the hybrid mode with a clo with close to 100 presentation delivered on the day one in the day to the Saturday at 30th May 2026 the day open with opened at 9:00 a.m.

[05:06:39]with a keynote address by Dr. Ishu Sharma the rank among the world top 2% scientist by Stanford University on the trustworthy air research framework followed by Dr. Shan Kumar Puja a professor from NIT Charundar delivered a keynote session on designing a mathematical model on human body and human physiology the two more parallel technical session throughout the morning completed by completed the program by uh 2:00 the conference highlights uh we have done around 180 paper presentation across 2 days 18 sessions across five parallel tracks 15 session chairs from the institutions across Karnataka and we have identified 18 best papers one paper per session the session was conducted in hybrid mode and even we had a presentation from India Taiwan USA and woman in one single platform our gratitude to the principal Dr. Santoro Mural and Dein Raj Raji Rajeshwar Karat for their guidance and support to our 15 session chairs thank you for your prime time your rigor and commitment to our four keynote speakers chief guest Dr. Abishek Apaji your contribution elevated the standard of this conference and also to the every student volunteer the faculty member of the organizing team.

[05:08:05]Two demanding days ran smoothly because of you to your presence your you revise stood before the panel to discuss your ideas. This is what the research looked like in practice. We are proud of everyone who presented the paper. I finally close the academic proceedings of ICTICS 2026. Thank you.

[05:08:35]Thank you sir for summarizing the key highlights of this conference. Now I request Dr. Santo Shia Munal, president of this session uh to distribute the momentto as a token of respect and gratitude to our session chair Mr. Manad TC on behalf of Amita Institute of Engineering and Management Sciences.

[05:08:58]Mr. Manganad TC please come to the stage on behalf of Amitan state of engineering and management sciences and organizing committee of international conference on trustworthy AI for intelligent computing systems 2026 we are very grateful to you sir it's now time to acknowledge the efforts of paper presenters and the participants. Now I request professor Madhusan to announce the participants list and I request Dr. Santo Y A Y A Y A Y A Y A Y A Y A Y A Y A Y Amul Nal President of this session and Dr. Rajeshwari Escar Devar chief guest of this session to kindly distribute the certificates.

[05:09:49]>> Thank you madam. It is time to announce best paper uh today's uh conference.

[05:09:57]Best paper goes to paper ID 64 engineers desk a web based solutions.

[05:10:09]Anyone uh from that team present name is not there only paper sir.

[05:10:17]Next names names Shivaputra Desai Raguir Praja College paper 67 a disability aware adaptive e-learning platform using AI and assistive technologies authors Amula is present.

[05:10:50]Next paper 25 road safety authors Jvita Bhumika Barat Gora Gat Gat.

[05:11:02]Anyone from the team? Yeah, please come.

[05:11:27]I request principal 17 sir please distribute the apps last minute Last one 78 25. Thank you sir. Paper ID 25 road safety. Authors are Shhat Shuumar from Amrit Institute of Engineering Management. Anyone from uh team please come.

[05:12:12]Paper 78 T478 UVM based functional verification from PS College of Engineering PS College of Engineering Mandia. Authors are Pavan and Sahana Raj.

[05:12:28]Next paper ID T243 realtime object detection authors are Sidhara Moit Murid Shastri Mitun Shinde Amritan Institute of Engineering and Management Sciences.

[05:12:57]Next T249 IoT integrated smart energy harvesting and automated integration system for sustainable agriculture from Amit instead of engineering management senses. Authors are Mahesh Matwati Mokshita Kajal and others.

[05:13:19]Next paper T1015 clock design generator using machine learning from BGS Institute of Technology. Authors are Bindu, Spandana, Oshita, Vina.

[05:13:31]Yeah. Next paper T60 IoT based IPG leakage. Authors are Kakala Patil Rakita from Shanabasala University, Kaluri.

[05:13:45]Next paper 046 T4 T4046.

[05:13:50]Enhance the efficiency of thin crystalline silicons solar cell. Authors are Mana then Sepkumar Ready and others from SRM Institute of Technology.

[05:14:08]Thank you sir.

[05:14:16]441 44. Next T4044 uh design simulation of RIS for smart wireless technology authors are Mjanata Manisha Harip Prasad and others from Nagajuna College of Engineering Technology Bengalur.

[05:14:34]Next paper ID 002 T1 food flow MIS intelligent food prediction authors are Priya and others from Institute of Technology Chennai.

[05:14:48]and T1056 towards trustworthy eye in disaster management biometric landscape. Authors are Sanjivkumar and others and uh uh school is ICFI tech school university.

[05:15:04]Next author paper ID T2075 adopt the behavior framework for insider threat prediction. Authors are Mamata Bhavan and others from BMS Institute of Technology, Cambridge, Cambridge Institute of Technology and other anyone present from there. Then T2018 Uni connect smart college app authors are Kavia Amutal Sigita from Amuta Institute of Engineering and Management Sciences.

[05:15:40]24062 autonomous for warehouse security atasar kavana punit kumar arun kumar and others ts college of engineering then track t 034 an AIdriven multisource authors are kumar bid simon Alexander and others sujan despande from amuta institute of engineering management sciences Then T4001 underwater biodirectional communication system using adaptive tech from Priti Sharma Puri and others global academic of technology.

[05:16:31]Thank you sir. Thank you sir.

[05:16:52]Once again, congratulations to all the participants. Let's give them a big round of applause, please.

[05:17:02]A conference is successful only when it creates value for its participants. We would now like to hear from you. May I request Sparti from Global Academy of Technology Bangalore to kindly come forward and share her experience of this conference.

[05:17:27]Good afternoon one and all present here.

[05:17:29]I'm Sparti from Global Academy of Technology. Uh firstly, I'm thankful for the entire management of uh Ames College for this great opportunity. Uh I think it's been almost one one and a half months from the process of application till today. the entire management has supported uh greatly and also there wasn't even a single drop down or glitch in the complete management and uh and also uh I'm thankful for the uh lecturers I have been I contacted them twice for uh during registration and all they're very helpful and also I'm thankful to the students uh from the core and subcore committee members they're very helpful today uh uh when whenever we ask something they're uh very helpful and uh uh I'm thankful for all the lecturers uh teaching and non-eing staffs also and and I'm thankful to the dignitaries who is sitting on uh the desk for their valuable time that they are dedicating for us. Thank you sir. Thank you one and all.

[05:18:37]Thank you for sharing your valuable experience and constructive feedback with us. Now I request Sesh P from Vidya College of Engineering, Mysuru to share his feedback with us.

[05:19:05]Hello, good afternoon everyone. Firstly, I would like to thank everyone. Thanks for the opportunity provided to me. Uh it's kind of a uh different vibe for me because I was the part of this college previously. Uh like I was passed out 2023 badge in this current college only.

[05:19:25]It was felt like happy for me while returning to college. I thought I I never thought that I'll uh come here and I'll uh step the stage again in on this particular auditorium or stage. It's a it was a good opportunity for me to experience once again the college moments, memories and those things. So each and every one of faculties I thank I will be uh forever thankful the principles are thank you thanks for this opportunity mainly and it was overall completely it was a good experience for me like the mail uh mail updates or whatever the WhatsApp group uh those updates whatever the changes those things are kindly it was um I mean not questionable it was uh can be updated do those things and whatever the organized from morning to evening completely it was um there was no um like it was proper organized thing so overall I'll conclude by saying thanks for this opportunity and uh thank you I'll conclude my words thank Thank you Sesh for say uh sharing your valuable words with us as an alumnist of this college. Your words carry special meaning for us. We are very grateful to you.

[05:20:58]We are privileged to have Dr. the Santos Muldan, principal of Amita Institute of Engineering and Management Sciences, who has been guiding us throughout this conference. Now I request him to present a presidential speech.

[05:21:19]>> Good afternoon everyone.

[05:21:22]Dignitaries Andas uh Dean Dr. RS car sir head of the department ECE Dr. Vesh Patin uh Dr. Mesh Panti professor CSC professor Shes head A and ML and all other faculties and dear students. So was a wonderful moment for the entire institute.

[05:21:45]Uh I congratulate all the faculty and students uh for this moment. So a big round of applause to the entire team.

[05:21:56]So the trustworthy AI for the intelligent computing system. So that is very much relevant to the present era.

[05:22:03]So that's why we selected uh this topic and I'm very happy that across the country from outside the country there's so many papers were presented and there's a good opportunity for all the students not just for sake of conducting conference. uh it's a good opportunity for all the faculty and students uh to understand what are different areas in which we can work then I congratulate all the uh participants or uh who have won the best paper award a big round of applause applause for all the participants and what is the takeaway from this conference especially for students and faculty we tutorial is the abstract. So please go through the abstracts and try to figure out what is your area of interest and in which you want to work and even you can contact them who have presented their work. You can take their guidance.

[05:23:04]So that's how as a technical community we can connect to each other. So already I think we are going ahead of time. So I wish once again uh all the best to the students. I'm always thankful to the entire team. So we have on different seven committees and uh faculty, non-eing staff, our students, everyone have worked very hard for a successful completion of this international conference and I wish and again I promise that coming ahead we'll have much more such technical events in the college. Thank you. Thank you very much.

[05:23:51]We are deeply grateful to our principal Dr. Santosia Muran for his thoughtprovoking address. Your constant support and leadership makes events like international conference and trustworthy AI for intelligent computing systems 2026 possible. Thank you sir.

[05:24:12]Today's offline participants are there here. I request professor Madus to announce their list and to give them a certificate and I request Dr. Santo to dispatch the certificates to them.

[05:24:31]Yeah. Today's uh offline session first session track ID T3007 T121 authors are Adita Vjala from BGS Institute of Technology. If present please come and collect a certificate.

[05:24:50]Shallini Sharma from Amita Institute of Engineering Management Sciences T129.

[05:25:06]T134, Bitkumar, Simon, Alexander, Susan, Pande, B.

[05:25:29]Yeah. T1 137 Harsh Ready Akshai Kumar from Amutan Institute of Engineering Management if present please come and collect the certificate.

[05:25:44]Sesh Dai Hamsi from BBC Mysore Shinas Dr. Naganandanda from Amita Institute of Engineering Management Instances P1 74 others are Nana Kushal and others PS College of Engineering India if present please come come and collect a certificate Rupashi from Amita Institute of Engineering Management Sciences T3005 Netraati Chaititra from AMS T3007.

[05:26:51]Abiga Priyadashini from Amita College.

[05:26:59]Next 39 from Amita College. Chinat Dashani Dashan Digshita.

[05:27:07]Nisa Nanda Priam Mokshita from Amita College.

[05:27:12]Sangeita Pri from Amita College if present please come.

[05:27:19]Jetanga from Amuta College Danish Gora from Amita College.

[05:27:43]Priti Sharma Sporty from Gat.

[05:28:10]Wait, man. Wait, man.

[05:28:25]Tapi man and others Bangalore Sani sirisha and Dr. hospital sir.

[05:29:05]Thank you, sir.

[05:29:11]Thank you sir. Once again congratulations to all the participants who have taken part in this conference.

[05:29:17]Gratitude is the memory of the heart. I now invite Dr. Mahhatesh Matapati convenor of this conference to present a lot of thanks.

[05:29:43]Good evening everyone.

[05:29:46]As we arrive at the successful conclusion of the international conference on trustworthy AI for intelligent computing system ICT IICS 2026.

[05:29:57]It is my great honor and privilege to propose the vote of thanks. On behalf of the organizing committee, I extend my heartfelt gratitude to all those who have contributed directly and indirectly to the successful success of this conference. First and foremost, we express our sincere gratitude to the management for their constant encouragement and support. We are especially thankful to our respected BES chairman uh Dr. Chantzer and our AMSGC chairman Masri Mahantesh Set Serji for their vision, guidance and commitment towards forecasting academic excellence and research culture. Their support has been a source of inspiration throughout the organization of the conference. I also extend our deepest appreciation to our respected uh respected principal sir Dr. Santos Munal Sers and Dean sir Rajesh Dr. Rajesas Kyra sir for their for their valuable guidance, encouragement and continuous support.

[05:31:09]Their leadership and motivation have played a pivotal role in making the conference a grand success. Our sincere thanks to our esteemed chief guest uh Dr. Abhisk Apazi uh are distinguished uh dictatories and keynote speakers, session chairs, invited experts for their sharing knowledge, expert and valuable insights. Their contributions have enriched this conference and inspires all participants. A special note of appreciation is to our conveners and co-conveners whose dedication uh mater planning and tireless efforts ensure and enrich the smooth conduct of uh every activity associated with conference. We gratefully acknowledge the contribution of the technical committee, editorial committee and the proceeding committee for their hard work in reviewing papers, maintaining academic quality and ensuring the successful publications of the conference proceedings. Our heartfelt thankful to public publicity committee for effort effectively promoting the conference and reaching a wide academic audience helping uh making this event a significant platform for their n for the knowledge exchange.

[05:32:39]We also sincerely appreciate the efforts of the stage committee, reception committee, hospitality committee, certificate committee, media committee and food committee. Their dedications, coordination and attention uh to detail the ensured every aspect of conference was organized professionally and efficiently. Without their hard work and commitment, a success the successful conduct of this conference wouldn't have been possible. A special word of thanks is to extended to all the faculty coordinators, student coordinators and non-eing staff who worked tirelessly behind the scenes. Their contributions, teamwork and willingness to take up responsibility at every stage were instrumental in ensuring a seamless execution of this conference. We are also grateful to the uh researchers, authors, industry professionals, academicians and participants who presented their work and actively engaged in discussions. Your valuable contributions made this conference a vibrant and intellectually exchanging the experiences. Finally, I extend my sincere thanks to each and everyone audience for your presence, participations and cooperations. We hope that the ideas exchanged, collaborations formed, and knowledge gained during the ICT AICS 2026 conference will continue to inspire innovations and research in the years ahead. With profound gratitude to everyone who contributed to the success of this conference, I consider this the word of thanks and thank you.

[05:34:23]Now the time for distributing uh the certificate appreciation certificates. I will call one by one the committees kindly come to the stage all the all together in the group. Uh I kindly request every the entire committee to come at once. First one publicity and media committee is starting and even the convenience to join their hands in uh thanking their effort and reaching to thousands of uh individuals across the country and across the globe.

[05:35:05]and we sincerely appreciate your effort in making this event a grand successful.

[05:35:13]Thank you. Thank you. Thank you.

[05:35:33]And the next what we have is editorial board of ISBN as proceed as editorial board.

[05:35:42]Follow Kumar. I please come on.

[05:36:00]See I wholeheartedly thank each and everyone standing on the deas for your effort from the day one in reviewing all the papers.

[05:36:13]Thank you very much.

[05:36:30]Thank you once again.

[05:36:32]Thank you. It was an amazing work that you have done. Thank you very much.

[05:36:36]>> Next conference proceedings and publication committee.

[05:36:39]>> Dr. Rita, Dr. Sanjan Prasad, Dr. Krishna, Professor Vijay Kumar, Professor Jay Shri, Vasant Kumari, Apita, Lakshita, Narati and Professor Kumar. Please command the deas we are really thankful for the day and night work to bring the proceedings uh released on time. Thank you very much for your and request and there is a special mention Dr. Sanjana is an editorial board of Atlantis press spring nature. Thank you for working for spring nature as well Dr. Sanjana.

[05:37:35]Thank you very much for making the move Kusama and Rashmi and Madam please come on this stage come on the doors please and we thank you very much for your effort that you have put in to make this event very beautiful. Thank you very much. Thank you.

[05:38:27]Okay. Huh.

[05:38:37]Thank you very much for your sincere efforts. Thank you.

[05:38:43]Next is registration committee registration committee Dr. Nagan Professor Prasa Professor Rakita Nan Shihut and Vive please come on the das and and even please come on the stage.

[05:39:14]How is it? And madam please.

[05:39:25]Thank you very much for make sure making sure the registration is streamless and uh without any errors. Thank you very much for your sincere efforts. Thank you.

[05:39:43]Next is hospitality committee.

[05:39:48]Uh professor Kabla BK, Professor Kabla, Mangjanatil, Rashmi, Axhata, Aishwara and Rashmi. Thank you very much for making sure that everybody is taken care. Thank you very much.

[05:40:07]Food committee is also joined.

[05:40:10]and uh hospitality food and transportation is what we expect. Dr. Pat, thank you very much for arranging the food for everybody.

[05:40:30]and also thank Mr. Malapa and Mr. Shindu for helping them in the food committee as well. Thank you.

[05:40:45]>> Thank you very much.

[05:40:47]Next is certificate committee.

[05:40:49]Certificate committee Mr. Masu Mangjanat and Vasimi. Three people have written close to 400 certificate. Thank you for your uh uh support.

[05:41:04]>> Thank you. Thank you. Thank you.

[05:41:08]>> Technical session coordinators session coordinators.

[05:41:15]>> Next is technical session coordinators committee Dr. Axita and Dr. Vinad. Thank you for making sure all these five power session run without any errors. Thank you for collecting all the reasons Dr. Thank you madam. Thank you sir for your efforts technical support. Sorry sir. Uh next student.

[05:41:51]>> Thank you madam. Thank you sir.

[05:41:53]>> Next is technical support committee.

[05:41:55]>> Technical support committee. Dr. Chaitan. Professor Ani Professor Chaitan Rakkesh Shinas Shilpa Aishwara please come on please.

[05:42:14]Thank you very much for making sure I'll call. Thank you very much for making sure all the sessions runs uh parally and troubleshooting all the problems that we came across with.

[05:42:38]Thank you very much.

[05:42:47]Thank you. Thank you for your efforts.

[05:42:49]Your efforts are greatly appreciated.

[05:42:52]Now what we have >> student last but not the least student coordinators committee see I more than what we have worked these team has worked day and night to make sure that whatever the technical things that we have seen over there in the days so I tal Danish adhes on the stage. Thank you very much for understanding and showing the maturity more than your age. We sincerely appreciate all your efforts. Thank you.

[05:43:44]Thank you. Thank you.

[05:44:13]And even thank you for working and thank you for working giving your food lunch and dinner for 2 days. Uh your effort is highly appreciated as well.

[05:44:26]Thank you. Thank you one and all. Any words?

[05:44:40]Good afternoon everybody.

[05:44:54]Okay, I want to congratulate each and every person teaching staff and non-eing staff for their dedicated hard work. Okay, we organized this conference within two 2 and 1/2 months maximum.

[05:45:11]Okay, that is highly appreciated.

[05:45:14]But we took nearly 6 months to one year to organize a conference. We took it as a challenge and we organized it and successfully completed the just today the professor from uh Rajar Rajeshwari Engineering College professor had come the conference is also going on there and he expressed not this professor the one professor had come he expressed that in their college also conference is going on but they have not received so much of papers from other states and within the states. He expressed really it is appreciated that you got a very good papers papers also very good and all the participants from all over the states they have participated and country uh really it is surprising you continue it's s like that blessed event okay this is how we have to next time we need to improve again and again okay there and there there are some flaws okay any thing anybody anything any work or anything has been hurted you okay I asked sorry for that okay uh next again this has to be continued for next year okay so every year minimum one or two conference should be organized thank you Thank you.

[05:46:54]>> Thank you sir.

[05:47:02]>> Good evening to one and all present here uh in preceding conference proceedings committee some of the names were missing hence I came here. Please uh don't uh consider this as an uh uh this thing. So I will call once again we will take a photo.

[05:47:19]Rohit from a IML department please come on the stage. Arpita Rashmi then Vijay Kumar some name were not there that's why I'm calling it was there you called all the names Rashmi names were there no come sir one second we'll take Rohit Arpita Rashmi Vijjkumar Kusuma Naga nan sir sir was not n was not therea jashri Nagapa Nagapa Patetti Nagap Patetti Rakshittita Ames Utam Utam from IC department certificates we will sit for you know join for photo session please. Wasant Kumar ready we wait we wait no we're not there please come.

[05:48:21]These are all the members involved in preparing the proceedings abstract proceedings.

[05:48:27]Thank you for all members.

[05:49:09]Thank you all. With this we come to the end of the international conference on trustworthy AI for intelligent computing systems 2026. Before we conclude I request everyone to stand in prayer for the national anthem.

[05:49:26]After national anthem, we have a tea break for refreshment.

[05:50:14]Never mind.

[05:50:26]Foreign speech. Foreign speech. Foreign speech.

[05:50:42]Foreign I request everyone to join.

[05:51:42]I request everyone to join for the group photo please. After that you can have a tea for the refreshment.