Day 3 | FREE Hands-On 3-Day Online AI Workshop : Build a Resume Screening Tool from Scratch

Añadido: 2026-05-05

[00:06:11]Okay. Good evening guys.

[00:06:15]And like every other session please do confirm am I audible to you and one more and is my screen visible to you? If you have any hiccups, please let me know. If I'm not audible or if I visiblely not >> okay Okay, please increase the Okay, that is fine. I will do your voice is very low.

[00:06:59]Okay, increase the volume. Good evening.

[00:07:02]Yes, sir. If everything is fine. Okay, we're good to go. Okay, now we'll have a small recap of what did we discussed in the last session. Then today we'll move to building this assume analyzer to and deploying in the AWS. Okay, before that we just have a small recap of the previous session. I need to do something disable annotation by this and clear. Okay, fine.

[00:07:38]Okay, I think in the last session we have discussed about we have discussed about uh while lang chain.

[00:07:52]Then we have discussed about after a while lang chain we have discussed about lang chain prompting that is in the form of messages in the form of messages. Then then we discussed about how to load chart models in lang.

[00:08:21]Once we have the discussion on how to load chat models and lang using lang chain then we have discussed about tools in the lang chain.

[00:08:33]Tools in the lang chain. After that we have we have just added a tool to the we have added a tool to the LLM added a tool to the LLM and because of it because LLM is interacting with the tool go on LLM go on interacting with the tool and getting the responses from the tool till based on the user query based on user query Okay, we have discussed it. It we have called it as a as or I can say asentic behavior.

[00:09:15]Okay, these are all things we have discussed in the lang chain. Okay, there are I have stated mainly three reasons for using the lang chain. In fact, there are many reasons available. Out of this, I have stated only three reasons to use the lang chain. Then the first one is the LLMs are stateless means LLMs don't have any memory.

[00:09:39]The second reason is with the LLM you cannot access the proprietary information or I can say if you want to build a chat port or if you want to build an application with the proprietary information you cannot use LLS.

[00:09:52]Okay. To build an application with proprietary information you need to use lang chain because lang chain has very good tools available to implement the rack systems. And the third one is if you want to know if you want to know updated information updated information LLM cannot provide the updated information. LLM can provide information only up to only up to the time they were trained. For example, if you look at the GPT5 last time it was trained on the August 2025. Then if you ask any question after August 2025 then you won't get the answer or you won't get the answer. It states that ask any question by time ask any question before the August 2025. If you want to access the updated information then LMS with LMS you cannot do it. Then for all these things LLMs are stateless. To add memory to add memory to add memory we need some tools and those tools are readily available with the lang chain. And to build an application with the proprietary information you need to build a rag system. For the rack system we need some tools. These tools are available in the lang chain. We can you can use the tools and build a rack system. In the same way, if you want to get the updated information, you need to use some tools and we have seen a large number of such tools are available. A large number of such tools are available in the tools and toolkits. With the help of these tools and toolkits, you can build an you can build an agent and agent can interact with the tool and agent can get the information required for These are the three main reasons why we are using this lang chain. Along with that there is one main reason for the lang chain that is lang chain is a unified a lang chain provides unified or is a unified framework irrespective of the LLM irrespective of the LLM the application code remains same application code remains Same. Okay. Only one line you need to change is the model name.

[00:12:34]Only one one line you need to change. If you want to switch from one LLM to another LLM, only one line you need to change that is model invocation or model name or model name. That's it.

[00:12:51]Okay. I have shown you the same in the last session at the time of calling a chat model. What I have done first I have built a chat model using Gemini Gemini 2.5 flashlight. Then just by changing one cell one cell information there is just invocation of the GR then the same code is useful.

[00:13:12]Then after discussion on this we have discussed about the we have discussed about writing the writing the prompts and for writing the prompts many procedures are many ways are possible like a chat prompt template is possible through the messages is possible now as far as our workshop is concerned okay okay we'll be using something called as the messages messages the first type of message is called as system message second type of message is called as human message.

[00:13:44]Through human message, we always ask a query to LLM. Through system message, we can fix the behavior of LM or we can give a role to the LLM. Fix the behavior of LLM or we can give a role to the LLM.

[00:14:05]Perfect.

[00:14:07]And there are two messages available.

[00:14:09]All I have shown you from where you can uh import this. After that we have discussed about we have discussed about the tools.

[00:14:18]Okay. In yesterday session we have used one search tool named as duck code.

[00:14:27]Duck go. With the help of this search tool we have found or we we we have found this very updated information.

[00:14:37]very updated information. Once we disc once we uh had our discuss once we had our discussion on this duck go then I have added this duck go to the llm with the help of something called as create agent in the l create agent in the lang chain then this create agent in the lang chain adds this llm to the tool add this llm to the tool lm two whenever you whenever you give a prompt to the LLM LLM will understand this prompt and LLM will go on interacting with the tool and get the required information once go on interacting with the tool in the loop go on interacting with the tool in the loop and once it gets the required information with respect to the user's query then it returns the response then it returns the aspects.

[00:15:41]These are all things we have discussed in the last and we have seen how to build a simple agent in the last session.

[00:15:58]Okay.

[00:16:02]Now in the previous session okay we have discussed about the we have discussed about the basics required to build the build the regime analyzer tool.

[00:16:14]Basics required to build the regime analyzer tool analyzer tool. Today we'll build this resume analyzer tool and we'll deploy this regime analyzer tool in the AWS E2 AWS.

[00:16:40]Okay. Now to build this regime analyzer tool. The first question we have is how to build how to now we have two options available to build this. One is just build the resume analyzer tool. just build the tool using lls or build the tool using a build the tool using as these are the two options we have one is lls and there is the aset but as far as regimes are concerned as far as regimes are concerned and as far as the as far as this Um recruitments are concerned.

[00:17:34]As far as recruitments are concerned for every 3 to four months, for every 3 to four months very new skills are comes into the market and new requirements are comes into the market.

[00:17:48]Nowadays it is very quickly the requirements are changed.

[00:17:52]new requirements are comes into the market and our LLMs are not really updating in every 3 months, every 3 to four months. Our LLMs are not updating in every 3 to four months.

[00:18:08]Not updating in every 3 to 4 months.

[00:18:13]And in order to analyze this regimes, In order to analyze these regimes, We need very updated information. Very updated market information.

[00:18:29]We need very updated market information.

[00:18:37]We need very updated market information.

[00:18:40]And in order to get this updated market information or very updated information, we are aware how to get this updated information with the help of tools and add these tools to the yield. Thereby it is possible to get the very updated information to analyze the ratios. Okay. I hope you you I think you have realized this fact. In fact, every 3 to four months a new roles comes into the picture. Very recently a new role comes into the market that is agent architect and agent engineers. very recently I think in in the India it is very recent in the developing country in the developer countries it is a bit a bit a bit preer but now it is very recent in India that is agent tech agent engineers okay every 3 to four months a very every every 3 to four months a new mark new requirements are new roles comes into the market but LLMs are not trained as often ls are trained for every new new LM versions are new model versions you obtain in every 1 hour, 1 year and 1 and a/2 year. Thereby one thing you can do is in order to access the new information and we have discussed in the last session. One thing you can do is the tools along with that along with that if you want if you are expecting the static responses.

[00:20:06]If you expect the static responses, whenever I say static responses, that is responses like just to summarize something.

[00:20:24]Uh if you are expecting a static tasks and the responses for this task like summarizing something like translating like correcting for all these tasks LLMs LLMs perform very well but but if you want to perform some reasoning and make decision if you want to perform some reasoning and make decisions based on the resonance.

[00:21:01]Make decisions based on the resonance.

[00:21:03]Then only one thing suitable is the agents.

[00:21:10]Okay. Agents can reason well and agents can make decisions by continuously interacting with the okay by continuously interacting with the tools.

[00:21:22]Okay, these are the two reasons why we are moving to the RA agents to build this application.

[00:21:28]Okay.

[00:21:30]Okay. Now we move to the agents to build this application. Whenever we move to whenever we decide uh to use agents and build an application then the first question arises is what is the agent architecture?

[00:21:45]Which agent architecture is suitable for your application?

[00:21:52]Which agent architecture is suitable for your application? As far as agent architectures are concerned, one agent architecture which is main which is quite suitable for this application or which is quite suitable for most of the applications is an agent architecture named as supervisor and subentets.

[00:22:14]Supervisor and subentets. Okay. If you look at this HR departments, okay, because mostly the regime analysis is performed by the HR departments and if you look at this HR departments, mainly the organization of HR departments looks like this. A head HR the person who makes the decisions followed by HR executives. First HR is given to HR executives.

[00:22:49]HR executives and based on this interaction between the HR executives with the head HR this analysis or decisions are made these decisions are made. Now in order to have a very similar architecture in order to have a very similar sort of decision making to have a very similar type of decision making to have a very similar type of decision making I'll be selecting an architecture named as an architecture named as a supervisor agent.

[00:23:30]A supervisor agent who makes the decisions by combining the informations obtained from By combining the informations obtained from the sub aents sub aent one sub aent two sub aent three and these sub aents collect the required information.

[00:23:56]These sub agents cut the required information I1, I2, I3 and based on this information collected by the sub agents, this supervisor agent makes a decision.

[00:24:10]Makes a decision. Now I want to really replicate the same type of architecture or same type of hierarchy followed in the HR departments. I want to replicate it. But as far as the sub agency is concerned, every sub aent is is specialized at some skill.

[00:24:28]Every sub aent is specialized at some skill. Okay. For example, my first sub agent is specialized at skill one and specialized skill two specialized skill three. Okay. Now before building this agent or this agent architecture first we'll discuss about what are the what are the skills and what is the workflow needed to analyze a ratio.

[00:24:58]What is the workflow needed to analyze a ratio?

[00:25:10]Okay. Whenever you want to analyze a resume, the first thing you need to understand, the first thing you need to have is the job description.

[00:25:21]Whenever I have a job description, one thing I can do is I can compare this job description with the resume.

[00:25:29]I can compare this job description with the resume. And in this comparison, in this comparison, if I find this rumé similar, if I find the rumé similar, then shortlisted.

[00:25:48]If this resume is not similar, not similar, then say rejected.

[00:26:01]Rejected. And if you want you can have some scores too. Okay, like the ATS scores. If there if the if there the similarity between the regime and the job description is high then it will assign a score like 90% then it is shortlisted. If the similarity is less it will assign a score like 40 40% then it will be rejected.

[00:26:25]Okay. This is the main idea of uh analyzing a ratio that is have a job description. Compare your job description with the rum. If the job description and resume are the skills, educa skills, education, experience and uh and the attributes are uh of of the candidate or characteristics of the candidate are sortable according to the job description then the person is then the person is shortlisted otherwise the person is rejected.

[00:26:57]Okay. Now, now in order to compare what are the criterias are what criter on on which criterians this comparison takes and what criterians and what criterians this comparison takes place.

[00:27:15]Okay. When I look at the criterians when I look at the criterians mostly on three basis this comparisons takes place. The first one is first one is the skills and education skills and education background. Second one is the professional experience or work experience.

[00:27:46]And the third one is a very important one that is the salary and the salary. Okay. What is the salary alignment?

[00:28:00]What is the salary alignment? These are the three things mostly people will look into while while shortlisting a resume.

[00:28:08]Okay. Now what I will do is what I will do is I'll assign a specialized agent for these three things. I'll assign my first agent my first specialized agent to take care of the skill and education.

[00:28:25]My second specialized agent to take care of the work experience and third specialized agent to take care of the salary alignment. Okay. Is the salary which is asking as is very is the salary which is demanded the salary which the person draws is on par with the on par with the market salary or not. Okay.

[00:28:50]These are the three things which really these are the three things which really most of the HR departments is focused on while shortlisting the ratio in comparison by compare by comparing with the job description. Now what I will do is because I have three main areas or three criterians found three criterions found to three criterians found to analyze the resume. Now what I will do is I'll assign I'll assign specialized agents to this. Thereby my architecture turns out to be a supervisor agent.

[00:29:26]A supervisor agent. This supervisor agent takes input from the takes input from the first agent who is taking care who is taking care of or first sub agent who is taking care of matching the skills and education matching the skills and education with the job description.

[00:29:50]Then the second specialized agent or second sub aent who is specialized yet at okay matching the work experience matching the work experience with the job description. And the third one is the another third sub aent who is specialized at matching the or finding the salary alignment.

[00:30:20]Finding the salary alignment for a particular role according to the job description. Now what I need to do is I need to create three sub aents. First sub agent task is to compare the skills and education of the candidate with respect to the job description according to the market or with the updated information. Second sub agent task is to compare the work experience of the candidate which is provided in the resume with the job description based on the terminology or based on this update based on the updated terminology which is available in the market. And the third subent task is to find out the what is the salary what is the salary range which is offering for this for for for the particular experience or for the particular uh profile in the market.

[00:31:16]Okay. What is the salary which is offering by the organizations for for the particular profile in the market.

[00:31:22]Okay. Once they collect these informations, all these informations are sent to this supervisor agent. Okay, that is agent one information, agent two information, agent three information. By consolidating this agent 1, 2, three and informations. Then supervisor agent make a decision whether to whether to shortlist whether to shortlist or to accept or whether to accept or to reject. This is the architecture for this is the this is the sortable architecture for the regum analyzer tool as far as the agents are concerned because we are using m we are using multiple agents and and creating interactions between the agents because we are using multiple agents multiple agents that is subent one sub aent two sub aent three are skills and educ a agent which is responsible for the skills and education comparison are skills and education match matching agent which who is responsible for the skills and work experience matching and agent responsible for the salary alignment matching and the fourth agent is the supervisor agent because we are using multiple agents and we have built the and we are building the interaction between these agents interaction between these agents this architecture is called as multi- aent architecture and very recently people started calling it as Agentic AI multi- aent architecture or agentic AI these agents within themsel they communicate and make a decision whenever I give whenever I give an input okay that is whenever user gives an input user gives an input in the form of resumeum and job description okay all sorts of interactions happen between these agents and if At the end these agents gives a decision that whether to shortlist the regime or to reject the regime. Okay.

[00:33:36]This this architecture is called as supervisor and subent architecture and this is the suitable architecture to to build this regime analyzer. This is about the architecture and the plan for the regime analyzer.

[00:33:53]Once okay once you once uh uh you are you once you are comfortable with this then we'll we'll move to the we'll move to the building this architecture already we have discussed about how to create an agent and we have discussed about how to add a tool then in order to build an agent we are aware of how to create an agent and how to add a tool that is now we need to create four four we need to decide four tools and create four agents and build interactions between them. That's it.

[00:34:25]Okay. Now in order to build interactions, one thing you need to aware is whenever the subent one whenever the subent I'll draw one more diagram.

[00:34:36]I'll draw one more diagram. This is the architecture supervisor and subent architecture and subent architecture an asentic AI architecture.

[00:35:02]Okay. In this whenever user gives an input in the form of in the form of resumeum and job description then all sort of interactions happen between the supervisor agent and the and the reporting sub is the reporting sub aent. This first sub agent is responsible for skills and education matching map skills and education matching. Second sub agent is for work experience. Third sub agent is to check for salary alignment.

[00:35:49]Okay. Once the skills and education matching agent done with their work, then the skills and education agent communicates his result to the his result to the supervisor agent. In the same way, work experience agent should communicate the result to the supervisor agent. This salary alignment agent should communicate result to the supervisor agent. Based on these results, result one from the skills agent, result two from the work experience agent and result three from the salary alignment agent. Based on these results, this supervisor agent makes the decision.

[00:36:35]Okay. Now this is these are the interactions. We need to make an interaction between the skills agent to the supervisor agent, work experience agent to the supervisor agent and salary alignment agent to the supervisor agent.

[00:36:48]Thereby thereby we can successfully build this architecture. Okay. Four agents and three interactions. Okay. And we need to build an interaction between supervisor agent and the three sub agents. Yeah. This is the architecture.

[00:37:01]Once this interactions is built then the based on the information from the these three sub agent supervisor agent can make a decision. This is the architecture to build the resumeum.

[00:37:14]Now already we have discussed about how to create an agent that we will see once. Okay. To create an agent.

[00:37:22]To create an agent. Already we have seen a function to create an agent in the lang chain. That is create agent with the help of create agent. You can create an agent. And if you want to add a tool, a lot of tools are available. A lot of tools are available in the uh in the lang chain uh in the lang chain. Tools and kits. Tools and toolkits. Tools and Toolkits library and in the Lchain library tools and toolkits sub package and if you want to add a tool we have seen the decorator at the rate tool to yeah to create a tool or to create a tool definition to create a tool now I'll be using this and first I'll start with the skill and experience agent then I'll build the then I'll build this work experience so skills and education agent then I'll build the work experience agent after that I'll build the first I'll build skills and education skills and education agent work experience agent then I'll build the third one as the salary alignment agent salary alignment agent And I'll combine all these agents information and I'll give these three agents information as a tool to the as a tool to the to the supervisor.

[00:39:00]Thereby the interactions are built successfully and you'll get the response. Okay. Now we start with the skills and education basically. Okay.

[00:39:11]But in order to build this agency, in order to get the response, you need an input. What is the input? A resume and job description. Okay, I have enough. I have created some rums and job descriptions.

[00:39:24]Some rums and job descriptions. And we'll be working with that reg rums and the job descriptions.

[00:39:31]We'll start.

[00:39:36]Okay, this is what this is the last session. Now we'll start with the today's session that is okay. Now to create an first we'll start with first agent named as an agent one. I'll call it a specialized agent one.

[00:39:54]Specialized agent one. Okay. Who is responsible for skills and skills and education match?

[00:40:08]education comparison okay between resume and the job description.

[00:40:18]Now in order to create an asset again I'll go to this lang chain I'll create I'll go to this lchain library.

[00:40:41]And if you want to access this, if you want to access the functions or classes in the lang chain library, then go to the docs. In the docs, as a second tab, you'll find the as a second set of tabs, you'll find this lang chain documentation that is deep agent lang chain and langraph documentation. And because we are using lang chain documentation, I'll go with this lang chain. And in the lang chain, if I want to create an agent, just in the in the first in the first page itself, I I can find a syntax. This is the aset. Yes.

[00:41:13]And this is the syntax used to create an aset from lang chain agents import. I'll copy this.

[00:41:24]Yes.

[00:41:27]Now okay from lang chain that agent and what is my first agent? It is the specialized agent one skills and education agent. I call it as a sked agent. Skills and education agent create agent. Then the model name and the tools. Model name and the tools. And I'm not I don't want to use this model name.

[00:41:56]I I don't want to use this any open. I want to use the Gemini. For the Gemini, how to give this model name? It is gemini sorry Google jennai.

[00:42:09]This is another way of giving this model name google jenna colon gemini 2.5 I want all we have discussed about this model flashlight.

[00:42:30]Okay. Now with the help of with the help of the Gemini 2.5 flashlight our LLM can interact with the with the tools but but one thing we need to do is first I need to import this Gemini to import the Gemini what I need to do I need to go to the I need to go to integrations in the integrations go to the Google in the Google go to the chat models In the chat models, I can find the syntax for importing the Gemini.

[00:43:06]Okay, this is the installing Gemini.

[00:43:13]Installing Gemini and after installing Oh, chain code is asking for register even that is fine. Okay.

[00:43:34]installing Gemini. After installing Gemini, then import the import the Google generate vi. You'll find the same. You'll find the statement here. You'll find this importing statement directly here. Just copy this.

[00:43:55]Once you copy this as the next step is okay in order to access this I need to provide an API key already already in the secret keys API key is available now I'll be using this secret keys API key and I will enable this Gemini okay the name is Gemini now in order to do that I need to import libraries import voice and from user data Sorry from google.c collab from google.c collab import user data then I can then I can do this first I want to get the Gemini key Gemini key from the secrets that is user data dot get of what is the name Gemini G M by Mi thereby thereby the key is stored in the Gemini once the key is stored I need to Assign this key to the environment v variable os dot environ of os dot environ of there are two ways one way is the google api key or you can go with either gemini api key assign gemini key to the environment variable assign gemini key to the environment variable Okay, instead what I will do is okay uh I'll share this collab link. Okay, just uh rather than typing, typing takes a lot of time. Just copy and paste. That would be quite useful for you.

[00:45:57]Yes. And I have I have pinned this message. Then everyone can access this.

[00:46:02]Use the use this and okay just copy copy and paste there way that way you can build very quickly along with me.

[00:46:13]Okay.

[00:46:15]Now, okay. Now I have given API API key to the environment variable. There no need to provide there no need to provide no need to provide API key whenever I'm using the whenever I whenever I instantiate the model. Okay. Now in the agent I have given the Gemini model and for in order to use this Gemini model I have given the Gemini key to the environment variable. Now the next question is tool. What would be the sortable tool for matching the skills and education. Okay. Whenever it is education the university names university way of dealing with the things are provided along with this whenever it is skills all the new skills and all these a long list of skills are provided to evaluate this now I need to now now I want to have a tool okay now I want to have a tool now if I want to have a tool then what I'll do is I'll go with this integrations in the lang chain integration S you can find the tools and toolkits. In the tools and toolkits a large number of tools are available. A large number of tools are available. Then we can select any one of the tools out of which if you if you are really interested the search tool and I have shown you one tool which is quite popular but we but you need to do some setup that is TV set. Okay. and and in the s tools even duct go works very well that we have experienced in the last session. Even duck go works very well that we have experienced in the last session. And what this duck go will do is this duck go will search whenever whenever lm ask any information dctor go search in the web and it come and it comes up with the better information in response to the query. If LLM is happy with the with that response then LLM will directly print the final response to the user or return the final response to the user. Otherwise LLM go on interact with the DG go tool and get the get the response which is suitable for the user query. This is how the agent works with the tool. This is how the LM tool interacts and this is how the agent works.

[00:48:40]Okay. Now for the skills and education for the skills and education I think no need of this web search we can directly go with the another popular tool here another popular tool in the lang chain is the Wikipedia we can directly go with the Wikipedia tool and we can perform this yes search for Wikipedia tool thereby directly you'll find two tools here one is wiki data and is Wikipedia now we are not Now we don't want any data wiki data. Now now go with we'll go with the Wikipedia. In order to use this Wikipedia just copy this install state.

[00:49:20]Copy the install statement and paste here. Okay. Now I want to have a tool and that tool is the Wikipedia.

[00:49:33]Install this Wikipedia.

[00:49:35]Then these are the import statements for the Wikipedia.

[00:49:46]These are the import statements for the Wikipedia to once it is installed.

[00:49:51]Okay. The problem is Lchain community is not installed. Okay. We'll install Lchain community.

[00:50:01]We install Lchain community or I can say updated version of Lchain community.

[00:50:10]Sometimes it ask for restart.

[00:50:24]Okay, we are very fortunate. It did not ask for restart. Okay, now import this Wikipedia.

[00:50:32]import this Wikipedia tools. Once we once we import the Wikipedia tools then this is the syntax available to create the tool.

[00:50:42]Yes. Now a Wikipedia tool is instantiated.

[00:50:50]Yes. Once we instantiate the Wikipedia tool, we can directly use this Wikipedia tool and get our web.

[00:50:58]I'll give here Wikipedia.

[00:51:01]This is the tool I want to use in order to match the skills and educational experience. But you need to communicate the same to the LLM. That is a tool is a tool is added to this agent. If you want you can use this tool and get the get the required information. This is this is one thing you need to communicate with the agent. Okay. If you don't communicate with the agent sorry LLM this is one thing that is a tool is added in the agent okay that is Google Gemini 2.5 flashlight a tool is added to the a tool is added to this agent that is Wikipedia tool if you need any if you want to extract any information then use this Wikipedia tool and get the information this is one thing you need to communicate with the LLM thereby LLM will be aware of it and LLM will use it Okay to do that in order to in order to provide that information to the LM what I will do is I'll go with the system system in the system pro I'll provide that information okay that is a Wikipedia tool a Wikipedia tool is added to is added. Okay.

[00:52:26]Is added.

[00:52:28]Uh okay. Rather than this first I need to fix the you are helpful HR executive.

[00:52:45]Use the Wikipedia tool.

[00:52:49]Use the Wikipedia tool and match the match the skills skills and educational information available in the resumeum educational information available in the regime.

[00:53:13]skills and educational information. I'll go to the next slide. Available in the resumeumé with the job description with the job description and and once you do this or once you perform this comparison once you perform this comparison then write the rest then write the result or return the result.

[00:53:43]Then return the return the response in following format.

[00:53:59]In following format the format is like this. The format what I want to choose is just percentage of match match in terms of percentage that is percentage and any missing skills.

[00:54:21]missing skills in the form of a list and just give some rating rating by 10. Okay, just respond in this way.

[00:54:42]Respond in this field.

[00:54:45]Yes, an agent is built. This is this that is whenever I provide whenever I provide resumeum and job description.

[00:54:57]Then model will understand this rum and job description and it ask tools to compare the compare the uh resumeum with the job description and this tool will give some response. When are whenever this tool gives some response then if this response is not really suitable for example tool gives some responses it is a good fit then model don't accept model want response in this model want respond specific format with the specific information only when model receives that information then this interaction will stop till then it go on interacting with the model it go on interacting with the model okay model will direct directly. Yes. Uh without without any human intervention model will go to the tool and get the information and it repeatedly interacting with the with the tool.

[00:55:55]Okay. In this in this repeated interaction there is no human human intervention model will make decisions.

[00:56:02]Okay. Whether this response is suitable for the user query or not. If it is not again go to the tool and get the information. Again if it is not go to the tool and get the information. This loop continues. This is called as this is called as agentic behavior and we call this as an aset. Okay. Now what I will do is what I'll do is I'll just take some information which is available in my notebook and based on that information I have some information available some reg information available with my name.

[00:56:35]I'll take this information and I'll try to use this information and get the response. This is the information. I have created a simple uh resumeum and job description in the form of text.

[00:56:53]In the form of text.

[00:56:56]Okay. Now we'll run this. How to run this? All we have seen response is equal to what is the name of the agent? sked agent dot invoke of then how to call this agent with the help of invoke and provide input in the form of human message in order to provide input in the form of human message first I need to import the human message from langchain dot messages import we don't need system message In fact system message okay and human message as a practice I have been doing it but we don't need any now we need to ask the you ask our query in the form of human message human message and content content is equal to analyze the analyze the analyze the I can use strings already we have discussed Test analyze the resume.

[00:58:10]R E S U M E. Okay, that is I have given this name analyze the resume with the analyze the resume with the given job description.

[00:58:39]given job description and in which variable it is stored job okay then I'll get the response what is the problem message human message invalid syntax that is fine. This should be in the form of a dictionary.

[00:59:09]After list it is a dictionary. Okay.

[00:59:20]Rum E S U M E and job_D job description.

[00:59:33]response.

[00:59:34]Why contents are required? It is provided already.

[00:59:40]Okay. Human message and recipes.

[00:59:50]Analyze the resume with the given job description.

[01:00:02]Wikipedia tool is given.

[01:00:05]Rumé is given.

[01:00:09]Okay, what I'll do is instead I'll go with the I'll go with the dock string and I'll provide the information manually.

[01:00:24]Analyze the resume with the job description with the given Job description.

[01:01:20]The prompt is given and once the prompt is given, commas are provided.

[01:01:29]After prompt already we have created an agent. Even agent is doing well. Then why should it gives this uh error?

[01:01:42]Now they have provided the return prompt.

[01:02:36]message. In the message, it is a human message and the human message.

[01:02:43]Okay, it should be contents.

[01:02:49]Okay, it should not be okay. And why is it asking this system prompt?

[01:03:37]I'll write it again.

[01:03:42]Response is equal to what's the response? It is sked agent sked agent dot invoke of messages in this human message.

[01:04:08]In the human message, it is a string.

[01:04:17]Analyze the given regime with the job description.

[01:04:41]Job description joes Okay.

[01:04:59]Perfect. It have analyzed.

[01:05:02]It have analyzed. Now, now I want this in the form of I want this. Okay. But it it seems like a mess. Now I want this in the form of a structured message for message in response of messages response of messages just message. Print message.p pretty bit thereby you'll get the you'll find the clear interaction between the you'll get the clear you'll find the clear interaction between the uh uh LLM and and the tool first human message analyze the given regime then AI message AI message calls the tool once it calls the tool then this is the tool message and it is it is uh it is just looking for all the information provided And these are the these are the different different costs with the tool message. And at the end at the end after discussion or after doing this or after comparison it come up with one result.

[01:06:12]What is the result? The match between the job description and the rum is 80%.

[01:06:18]And one skill which is missing is the langraph and the rating for the rum is the 8 bit.

[01:06:26]Yes, this is my skill and educational agent response. Now in the next step, what I will do is I'll send this response to the supervisor.

[01:06:38]I'll send this response to the supervisor.

[01:06:42]Okay. But one one problem for all these things is my my prompt is not really a very good prompt or my prompt is not a standardized prompt. I need to ask this in the very very very detailed way or very uh standard way uh in a very precise way as we discussed in the last session if a prompt is precise you get your response as a precise response now what I will do is instead of asking prompt in this way instead of asking prompt in this way now I I I have already one prompt available I have already one prompt available uh that is a very stand very precise prompt I'll go with this precise prompt. I think hope you get this. I hope you get this idea how to ask this. Okay, this is the prompt.

[01:07:28]This is one prompt.

[01:07:33]This is the prompt. Now I'll replace this with the with a precise prompt what I have written to perform the same. If you if you if you study this prompt, both the prompts are same. There is no much difference in the prompts. Only thing is I have written in the very ne very precise way here. That's it. Now I'll run this agent regime and response.

[01:08:12]Okay. Because there are because it needs to take many costs. Yes. Starting with the human message then directly the match is 75% around around 80 and the missing is the cubernetes 8 by 10 okay whenever I provide the precise precisely Kubernetes missing is found out along with that this is a short summary about the skills and skills and the uh educational information skills and the educational information.

[01:08:43]Okay. This is my first specialized agent called as skills and education. Skills and whose intention is or who specialized at comparing or matching the skills and education. I in the same way I need to create two more agents. For creating two more agents, what I will do is I'll go with the same syntax.

[01:09:05]I'll copy this.

[01:09:08]The next agent is the specialized agent two.

[01:09:11]Specialized agent two and the intention of this specialized agent 2 is just compare the compare work experience.

[01:09:25]Okay. Now hold now for this work experience for this work experience this standard Wikipedia is not enough. In order to compare the work experiences, in order to check the companies and all, in order to check the given company names, in order to compare the work experiences and get the required information, okay, this Wikipedia is not enough. It needs a search tool and we are aware of one search tool which works very well that is called as duck go search. We'll go to the duck go and we import the data for the search to okay because in order to compare this work experience it needs to search the web and find out the information about the work experience.

[01:10:10]For that it needs a better tool than this Wikipedia. Thereby I'm installing this duck.

[01:10:21]Now import the duck.

[01:10:24]Install the duct go. Then import the duck. Here it is available. This is the import statement.

[01:10:32]Import the duct.

[01:10:36]Yes. Now d is imported with a name search.

[01:10:56]Okay. Now instead of Wikipedia now I I'll change the I'll change the agent name that is uh experience work experience asent work experience asent. Now instead of Wikipedia because in order to check the work experience in Wikipedia it is not possible. Okay. I need to I need to search in the web thereby I have used this duct. Now as far as prompt is concerned from the work experience again I need to ask for the match to ask for the match. Okay if I write some prompt that prompt may not be suitable thereby I'll go with the prompt which is available with me. I'll go with the prompt which is available with me. This is the prompt.

[01:11:47]This is the prompt. A resume evaluates the assistance of work experience and after evaluating the work experience provided in the or after analyzing the work experience provided in the resume and the job description at the end at the end print the return the following.

[01:12:03]What is the fit between these two? What is what are the list of companies the person worked and what are the roles the person worked and what is the rate. This is the agent.

[01:12:15]Okay. Once I have an agent available, once I once I run this agent, already the input is given here. Now I need to call the agent.

[01:12:27]An agent is instantiated. Now and the prompt is also prompt is al also given now.

[01:12:36]Now just look at the response but it is not the SKD agent. It is work experience isn't W experience work experience is yes both are analyzed and and the fitt is 90% and this is one company where I worked previously a kohir a kohir health company then okay these are the rules what I have served for then the rating is nine rating is nine Now we have created the second specialized agent too that is the work experience agent. Now the last one is the salary agent. Last one is the salary agent. Now to build the salary agent again. Okay, it should be WexP agent not Wex agent and okay anyway it is fine.

[01:13:38]Now I want to build the third specialized agent called as salary agent with the same with the similar syntax with the similar syntax salary alignment agent. Special agent three.

[01:13:55]Salary alignment agent.

[01:13:59]Salary alignment agent and even for the salary alignment agent.

[01:14:11]Okay. I need to search in the internet about the about the salaries of this experienced person. salaries of five plus years experienced person who has worked or who has served in some in in so levels with the so and so designations in order in order to find that I need to search them I need to search in the internet okay again I need to search because I need to search in the internet I have used this searches I I want to use this search to now here here I need to write a prompt okay the prompt is nothing but just analyze this and return return the return the indicator return the salary range in order to do that I have prompt available I'll just use the same prompt okay instead of writing the prompts now instead of writing these prompts now it takes a lot of time rather than that I have all the prompts available and we know the intention what exactly I we need to Okay. Then just analyze the just anal just find out the find out the profile of the profile of the person in the resume and just give the indicative salary. Give the indicative salary. This is what I'm asking.

[01:15:36]Salary agent. Okay. I need to write here salary agent.

[01:15:44]salary agent and this is the prompt for the salary agent. Once I have this salary agent then what I need to do I need to invoke the salary agent by providing the resume and job description by providing the resume and job description. It is not work experience is it? It is salary agent.

[01:16:47]Maybe maybe it is taking a lot of calls to cars. That's the reason why it's going uh running.

[01:17:29]I stop execution.

[01:17:33]There is a reason for running this capsu for the for the last time that is sometimes this tool when because I am using this set tool set tool is doing API calls sometimes this API calls uh API requests are not processed or API requests got gets the errors and API requests are error out sometimes maybe there's a reason why it takes a good amount of time I need to stop it otherwise else it will be delayed.

[01:18:06]Okay, restart the run time.

[01:18:14]Okay, once I restart the runtime, I need to start from I need to do it from the start.

[01:18:48]Okay, leave it. Uh, from this point onwards we have started today.

[01:18:54]I'll run this.

[01:19:04]Okay.

[01:19:06]Now, now once these three are completed, the next step would be I need to discuss a bit about the next step. The next step would be okay give this results of these three assets that is skills and education is it, work experience is it and the last one is this salary alignment is it. I need to give these three responses.

[01:19:35]as an input to the as an input to the supervisor isn't supervisor but okay and again this if you look at this supervisor agent again supervisor agent needs tools as an input not the agents as an needs tools as int.

[01:20:07]Okay, if you look at the agent syntax, what is the syntax? It is create agent.

[01:20:14]Create agent and model name tools.

[01:20:20]Okay, every agent needs tools as an input. But we have agents available.

[01:20:27]Every agent needs just tools as an input. Now what I need to do is I need to convert the responses of these agents. The responses of salary agent, the response of the work experience agent, the response of the skill and education agent. Now we need to convert these responses to as a tools with the help of a decorator named as the rate to with the help of a decorator named as at the rate to I need to convert the responses of this agent to the tools.

[01:21:01]Once I convert the responses of these agents as a tools then these tools can be given as an input to the supervisor agent. Okay. As we have seen in the syntax of create agent, agent only takes tools as an input, not the agents as an input. Now what I need to do is I need to convert the response of these agents as a tools and I need to give the tools to the to the supervisor. Yes.

[01:21:30]Okay. Now after uh now by searching through the web by searching through the web it found it found that the salary range is 8.75 to 12 LPA and the height is hike is not applicable as far as the as far as this market is as far as the market is concerned and the confidence of this result 8.75 to 12 LPA and the hike is not applicable is medium. Okay, this is the response obtained from the salary agent. Now we got the responses from these three agents. Once we got the responses, now we need to convert these responses into the tools. How to convert? In order to convert that all we have seen in the last session lang chain dot tools import tool.

[01:22:20]Now create a function with the define define.

[01:22:26]The function is call sked agent.

[01:22:30]Call sked agent and inputs are rum as an input and job description as an input. By taking the rum and job descriptions as an input.

[01:22:45]Okay. call this SK agent and I need to provide the information otherwise agent don't know what exactly the tool is okay this is this tool is used to analyze the analyze the resume analyze the resume in terms of in terms of skills and education skills and education and here make the car okay in order to do that rather than doing all these things manually what I'll do is I have already a function available for this I'll just use this function okay but the idea is same idea is a supervisor agent cannot take an agent as an input it it takes only tool as in it thereby what I have I have created a function and I have given the response statement. I have given this response statement here. That's it. Okay. All they have written many response statements. See this calling response is equal to salary agent invoke of messages. The same is given in the form of a function. That's it. The same is given in the form of a function. Okay.

[01:24:06]Instead of instead of getting this output manually, I have just given all these things in the form of a function.

[01:24:15]Yes, I have it I have created a tool for the skill and education matcher. In the tool what I have given I have just done the response state or the agent calling sk agent.

[01:24:31]In the same way for the remaining two tools also for the remaining two tools that is work experience is it I need a tool work experience is it need a tool it is work experience is it double you experience a two and and salary matching agent because agent cannot take another agent as an input.

[01:25:17]Thereby it needs tools.

[01:25:19]It is a tool for the salary major agent.

[01:25:22]Okay. In these tools what I have done is I have just given the invoke statements.

[01:25:27]That's it. just given the invoke statements. Previously I have given this invoke statements individually in the cells. Now I have written in the form of the tools. Now once the tools are ready, now I can create a supervisor agent. How to create a supervisor agent? Supervisor agent is equal to create agent.

[01:25:50]Create agent of model.

[01:25:54]As we discussed Gemini model we are going to use Google GI Gemini 2.5 flash light this is the model along with that I want to use some tools what are the tools here the tools are here the tools are This first tool is the call skill education matcher.

[01:26:29]This is the tool because I cannot provide agent as an input to the another agent. I have just converted the response of an agent in the form of a function and I decorated with the tool work experience match and salary match.

[01:26:56]These are the tools. Now once I given the tool tools then system prompt there is I need to communicate that the tools are added here in order to analyze the resumeum use these tools. This is what I need to provide.

[01:27:15]Okay. Now I'll take the prompt. I'll take the precise prompt what I have written for this. Yes, this is the precise prompt. Use the tools. Use the three tools given for you and from the tools get the information. From the tools get the information and decide whether the resume is accepted or rejected.

[01:27:44]Okay. Comma is missing.

[01:27:49]Yes. Supervisor agent.

[01:27:52]Okay. Now my agent architecture is built. I have built a supervisor agent and I have given the I have given the sub aents sub aents responses as an input to the supervisor is it thereby the necessary corrections are made and the agents are built okay now my architecture is ready once my architecture is ready now now now I'll I'll go for this I'll go for this calling this supervisor agent calling the supervisor agent and asks whe and ask whether this person is whether this person is sort whether this person is accepted or whether this person resume is accepted or recepted.

[01:28:38]This is this is how you can call supervisor agent invoke of messages analyze the resume with the job description and print the response.

[01:28:54]Yes.

[01:28:58]Okay. These are the different tool cards. After different tool cards.

[01:29:04]Okay. After different tool cards, uh first it calls the skill education m skill and education matter. Then it calls the experience matter. After that it calls the salary m.

[01:29:19]But it is not generated. Well, okay.

[01:29:22]Sometimes it happens. I'll show you a way to get rid of this. Sometimes it happens.

[01:29:29]Okay.

[01:29:31]Yes. First it calls this skill and education matcher, experience matcher and salary matcher. At the end it gives an decision that the candidate is a strong match for this role based on the scores provided by the skill and education experience and the salary then the decision is up is up. This is how this multi- aent this multi- aent tool works. Okay.

[01:30:04]This is how this multi-agent tool works.

[01:30:10]And this is the architecture of this multi- aent or this is the this is the rough code or a a simple code to build this regime analyzer too. But if you directly push this to the deployment environment it may raise some errors. In order to get rid of that I have made some modifications. Now first I'll show the modifications then we'll push this code for the deployment.

[01:30:36]Okay.

[01:30:38]I'll show this first. This is the code I have mod. This is the mod. This is the aligned code. This is the aligned code.

[01:30:46]These are the libraries I have used to I have used to I have imported. Then Gemini API key then first skill and education agent. Skill and education matching agent. But now look at this middleware. These are the new lines.

[01:31:04]Just by going through these lines you can understand what exactly is this.

[01:31:08]That is sometimes there is a possibility that your tool API call gets paid gets error. In such cases in order to in order to get the response from the tool I have given the maximum retries as three. That is whenever your tool whenever your tool is failed to communicate whenever your to your API is failed to communicate with the tool your tool API is failed to communicate with the tool then just try three times then respond it as an error otherwise do in the same way sometimes there are some errors with the model that we have seen already for the first time I run this application I got I got an error or the response is not really generated uh correctly in such cases is this model will be model will be I have provided the three retries for the model. Okay, the same prompt I have used.

[01:32:03]Next one is the experience agent. Even for the experience agent I have added this. Okay. When whenever there is a model whenever there is a APA calling error with the model just retry for two times API calling for the tool retry for three times for the salary asset. Whenever there is a API calling error with the tool retry for three times model retry for two times.

[01:32:30]These are the tool definitions. The same tool definitions I'm using here. The same tool definitions.

[01:32:36]And at the end I have just at the end I have formatted the output. I have formatted the output in the by using by using the pidantic model. Okay. Whenever I format the output by using pyantic model this can be further used to the API class. This can be further used to the API call. I have just formatted with the pedantic. Okay. If you don't format also doesn't makes any difference. Okay.

[01:33:01]only for the sake of better output and only for the sake of communicating this to the next downstream stages I have formatted the output. Then I have created the supervised result. Even for the supervised result I have added this retries and this is the same prompt we have seen. Okay. Now after this I have written a piece of code to load the document because every time it is every time copying the content from the resume and provided provide this content as an input it is quite tough. Most of the times rums and job descriptions are available in the form of the the form of a PDF documents. Then I have written a piece of code to load the PDF documents.

[01:33:45]To load the PDF documents and this is the piece of code to create the to create the front end with the help of streaml.

[01:33:56]Now once I have this code now what I'll do is first I'll run in the local then I'll push the CM to the AWS.

[01:34:05]First I run in the local stream lit run py what's the problem options okay this file is available in the music folder okay I'll run this in the locker Yes, it takes some time to load.

[01:35:22]It is taking a lot of time. Now what I'll do is I'll go back. I'll stop this uh execution with Ctrl C and I'll run this again.

[01:35:43]Yes, I'll run this again.

[01:36:13]Okay. In the meantime, it is loading.

[01:36:15]Now, what we'll do is ah yes, this is the whenever I provide resumeum in the form of PDF and job description in the form of PDF, it will analyze and give the rest. I'll show browse files. I'll go to the downloads. In the downloads, I have some rums and uh some rums and job descriptions are available. I'll go with the rum one, a data scientist resume.

[01:36:39]Open and I'll go with the browse. I'll go with the job description. Job description of the one once I load this regime job description.

[01:36:57]Now it will then this evaluate will be displayed. Just click on the evaluate.

[01:37:06]Okay. I'll share this code to you. I'll share this code.

[01:37:10]Click on the evaluate.

[01:37:28]Okay.

[01:37:33]The reception Browse files.

[01:38:00]Rumé 4 open and job description. Job description 4 open.

[01:38:08]Once these two are loaded then evaluate will be displayed. Once I get the evaluate, if I click on the evaluate, okay, then it will find the result.

[01:38:23]and it will analyze and get there is a reason I think I have deleted few pieces of code.

[01:38:37]Okay.

[01:38:43]No, in fact it is not.

[01:38:47]I'll save this.

[01:38:50]Save.

[01:38:52]Control C.

[01:39:16]Okay. In the meantime, what we'll do is we'll go with this AWS. Okay, we'll go with the AWS deployment.

[01:39:26]Yes, this is my AWS account. Now, we'll go we'll go for a quick EC2 deployment.

[01:39:32]For a quick EC2 deployment, I'll show you the steps. I'll show you the steps of how to deploy this in EC2.

[01:39:38]Okay, in order to deploy this in the EC2, what I need to do is first you need to create an environment. Just open open AWS and in the AWS go to the tools. Here in the tools you'll find EC2. Otherwise click on the EC2.

[01:39:55]You'll find the EC2. Just click on this EC2. Once you open this EC2 it will ask for instances.

[01:40:04]Go to the instances and there are no instances available.

[01:40:08]Now I want to launch a new instance.

[01:40:11]Just click on the launch instance.

[01:40:15]Once you click on this launch instance, it will ask you for it will ask you for any instance name and my instance name is the resume analyzer.

[01:40:31]This is my instance name. Now, now I want to deploy this in the Ubuntu environment for the sake of uh in order to get rid of the errors Ubuntu environment. Just select the Ubuntu environment. After selecting the Ubuntu environment, the next step you need to do is you need to search for you need to create a key pair. It's like a password. You need to create a key pair because I'm using a laptop with uh Windows 10. Then I can go with this PM file.

[01:41:04]Create a key pair. I need to provide the name. Provide the name is ratio.

[01:41:11]Create a key pair.

[01:41:14]Okay. Resumep a key pair is a key value pair is created. It is like the it is like the uh username and password.

[01:41:24]Once it is created then we'll go to the launch steps.

[01:41:43]Okay, I'll share this code files. Okay, now the instance is launched. We'll go to the instance. Yes, instance is launched. Now once the instance is launched, one thing you need to make sure is what are the security what are the what are the security settings or security groups. Just go to the security groups.

[01:42:08]uh go to the security groups and create a security group.

[01:42:14]Okay, prior to that what I want to show you is instead of going to the security group I want to show you the security settings first regime analyzer tool and security.

[01:42:33]Yes, in the security settings we need we we most we are mostly interested two things. One is the inbound rules, outbound rules. Outbound rules is from the EC2 what are what are all permissions we have we have to access the outside tools those are called as outbound to outbound rules. For the outbound rules for the outbound rules we have I think we have all the permissions available. Now look at the inbound rules. In case of inbound rules, we have very limited permissions available.

[01:43:03]There is TCP very limited permissions available. Now what I want to do is I want to go to the security groups and create a new security group. Create a new security group. Before creating a new security group, I want to delete the existing security groups. Delete this.

[01:43:24]I want to create a new security group which can provide access to which can provide access from any inbound to any outbound that is uh you can access from anywhere and you can access anything outside in the internet. In order to do that already outbound rule is fixed here. We can we can you can access anything from the outbound rules and from the inbound rules. What I what we need to do is we need to add we need to add just all traffic. Whenever you provide all traffic this gives an indication that you can access anything.

[01:43:58]You can access anything from the EC.

[01:44:01]Sorry, you can from anywhere you can access this EC2 from anywhere. From anywhere you can access this EC2 just add the rule add the once you add the rule then okay then create a security group okay before creating a security group I need to provide name to the security group I'll provide name as resume to resume.

[01:44:34]Resume okay only thing you need to modify is from to the inbound are you cannot with the pre with the previous inbound rule you cannot access this EC2 from anywhere. Okay. Now what I have done is I have changed this I have changed the inbound rule to I want to access this EC2 from anywhere. Okay. Once I have modified this security groups, then what I need to do is again go to the instances and I make my EC2 instance.

[01:45:05]Okay. And I make my EC2 instance update to the new security. In order to do that, just go to the select your EC2 instance and go to the actions. In the actions, go to the security and in the security go to change security groups and you add a security group. What is the name given? Resume to security group. Yes, this security group and save this setting.

[01:45:34]Then a new security group is added.

[01:45:37]Thereby you can access this EC2 instance from anywhere. Previously it is not. So now now you can access this sec this EC2 instance from anywhere.

[01:45:47]Now once my once my EC2 instance is built under security and and given enough necessary security group and necessary security rules. Now I need to now I can I can connect this to the Ubunt in order to connect this to the Ubuntu.

[01:46:02]Yes, I have an I have this available or I have some uh address given. Now copy this address. Open the command prompt.

[01:46:16]Open the command prompt and in the command prompt before uh before start before deploying this I need to push the code I need to push this code file to the Ubuntu or AWS computer or AWS server in order to push that I need to do the secure copy secure copy minus R - I what I want to copy I want to copy from okay before copy Before copying I need to provide the certificate.

[01:46:52]Uh this is the certificate. Copy this and paste it in the music because I want to access this through the music. Go to the music.

[01:47:06]From the music start copy secure copy - r - i. Now provide the provide the set this certificate or provide this uh uh key value pair. What is the name? Resume.pm.

[01:47:25]It is like a uh passport resume.pm resume.pm. Now just copy now secure copy the files from computer to the EC2. Now what is the name of folder available here? I have I have kept all these code files in a folder named as resume analyzer 2. Now just take this resume analyzer tool and paste it to the and paste it to the to this Ubuntu address or to this server address. This is the server address.

[01:48:08]Paste it to the server address.

[01:48:13]This is the server address and in the server address go to the root folder and in the root folder paste this.

[01:48:22]This is the root folder. Go to the root folder and in the root folder paste this. It will ask you for some permissions by yes.

[01:48:34]Yes.

[01:48:43]Failed to upload directory rat to Y.

[01:48:47]Okay. Now what we'll do is we'll again go back secure copy - r - i just I have provided this uh uh pm then copy this rat folder folder to the Ubuntu.

[01:49:11]No such file directory failed to upload is it so we are moving from the music in the music secure copy - R - I resumepm file folder name is a and paste it to the this is the address Ubuntu DC2. This is the address of my server. Paste it to this server.

[01:50:10]Okay. Till then start remote no such fire directory.

[01:50:21]Fine. That is fine.

[01:50:24]What I need to do is I need to check for the settings actually.

[01:50:46]Okay. A small problem here.

[01:50:50]Root folder and yes.

[01:50:57]Yes. Okay. I have provided three files.

[01:51:00]One is the requirements. Second one is the regime analyzer tool and product.

[01:51:06]Those are those are available with the Ubuntu. Now we now we will log to our server. In order to log into our server, I need to go to this AWS and this is the address of the server. Copy the address.

[01:51:19]Copy the address. Go to the command prompt and log to your go to your server.

[01:51:27]Yes. Now the server is opened. Once the server is open, now what I need to do is this is like a new computer. New computer with 6.71 GB of memory. Now I need to install everything to install this. First I will install first I'll update this. Update the OS pseudoapp update.

[01:51:53]I need to install everything. This is like a new computer. I need to install everything. First I have updated this Ubuntu.

[01:52:00]Once after updating the Ubuntu now what I'll do is I'll install the Python then I'll install the Python pip install the Python sudo ad install Python 3.

[01:52:21]Yes, Python is installed. Once Python is installed then then then pip install. Sorry, not pip install in the Python 3. Sudo apt install.

[01:52:41]Install Python 3 - pip.

[01:52:52]Now once pip is installed, you can install whatever library you want in this in this server.

[01:53:10]Okay. Now once PP is installed, now what I'll do is I want to install the virtual environment.

[01:53:21]I'll create a virtual environment and in this virtual environment I'll run this application.

[01:53:27]Yes. Once virtual environment is installed, I'll create a virtual environment. Python 3 - m create a virtual environment and name of the virtual environment is my environment.

[01:53:47]Your virtual environment is created.

[01:53:49]Once creating the virtual environment, now I'll get into this virtual environment. How to get into this? Just use the source.

[01:53:59]Source.

[01:54:02]Okay. Not directly from the source source. In the source root folder from in the root folder, my environment is available. In that environment, go to the bin. in the bin just activate it.

[01:54:28]Okay, now we'll just check this environment. Yes, my environment is available. Once my environment is available, I can go for this my source my environment bin activate. Okay, just I'll keep this aside thereby without mistakes I can do it. source my environment bin yet. Yes, I got into this virtual environment. Once I get into the virtual environment, before running my file, I need to install the libraries because I have requirements file available. I'll install this libraries.

[01:55:12]pip install minus r - r requirements dot text q u i requ requirements dot text okay spelling mistake r e q u u i r e m e ns txt.

[01:55:44]Hi Joe. Okay, problem is I need to get into this folder.

[01:55:52]Now pip install requirements. Now these are the different libraries which are needed to run our file. All these libraries will be installed.

[01:56:03]Once these libraries are installed.

[01:56:24]Once these libraries are installed thereby we can run our we can run our Python file. Once we run our Python file, it will give the it will give the IP. It will give the external IP or an IP which can be accessed by anyone in the world and with that IP anyone can use it and anyone can text it. Thereby we can or in a very easy way I can say thereby we can serve our application. Yes, once all these requirements are installed, now I'll run the Python file in the streaml in the stream with streaml str.

[01:57:11]Yes, now I got this external link. Okay, with this external link, anyone can use this. Okay, I'm just posting it through the chat box. If you are interested, you can test this.

[01:57:23]Okay, anyone can use this.

[01:57:26]And along with this, I'll go to the Okay, it is the local one. I'll go to this Okay, this is the resume analyzer tool.

[01:57:54]Now I'll give a resume uh resume five open and job description five open.

[01:58:09]Once these two are loaded, I'll get the evaluate. Once I click on the evaluate in fact I should get the response okay maybe because of my APA class I'm not getting what I'll do is I I'll share this resumeums to you you try to you try to test this by giving Please tell us everything is pretty fine. In fact, I haven't tested this files are given evaluated.

[01:59:15]Okay, just open this link uh open this link and provide the resumeums what I have given to you and test this. Okay, it should work. In fact, it should work. I have thoroughly tested this. It should work.

[01:59:42]Okay.

[01:59:44]Why unable to open? It's I have provided an external link. Anyone can open this.

[01:59:49]I'll give this again.

[02:00:34]Yes. Is it working? Okay. If it is working, it is fine. Fine. Okay. I don't know for me. PDFs are not available. It is working. Thank you. Thank you. Thank you. Thank you for confirming. In fact, I have thoroughly I have thoroughly used it and I have thoroughly tested it before the session.

[02:00:52]And it must work in fact for me I don't know why it must working perfect I will provide I'll provide this code files I'll provide this generated image too But I don't know why it is not working for me not working. Why P for you? In fact, P file is not needed to run this. In fact, I'll share this. I I'll share this file. I'll share this files. Okay, I'll try to I'll I'll make I'll make one more try once.

[02:01:52]Job description open.

[02:01:59]Browse.

[02:02:02]Resume. Open.

[02:02:06]Uh job description. Okay.

[02:02:17]Evaluate.

[02:02:21]Oh, refresh. Okay.

[02:02:37]Okay. Okay. That's good. That's good.

[02:02:43]Okay. key error structure in response.

[02:02:50]You should not get this in fact.

[02:03:39]Okay, I'll share I'll share that. I'll cop I I'll I'll prepare this as a notes and I'll send I'll send this uh how to run how to uh what are the guidelines or steps to deploy in AWS. I'll send that.

[02:03:55]Okay. And this is about this resume analyzer tool. Very unfortunate thing is it is not working in my PC and in with my with this connection. I don't know why uh this is the very unfortunate thing. If you if if it worked then it would be quite good. U and this is about this 3 days workshop on the regime analyzer tool. We have successfully built a resume analyzer agent and uh we have successfully deployed that agent in the AWS.

[02:04:24]Now I'll share this. I'll share the attendance link chat box.

[02:05:53]Okay, I have shared the link. Just fill the fill the Google form along with that. Now I'll share this uh scanner too.

[02:06:08]Yes, I have shared the scanner also.

[02:06:10]Just scan and fill the I share the QR code too and soon you will receive the certificates and I'll share this notes and code too.

[02:07:41]Okay, you'll get certificates soon.

[02:07:42]You'll get certificates. It will take some time to generate your certificates.

[02:07:46]You'll get your certificates.

[02:08:56]You'll surely get it. Don't bother about it.

[02:09:00]Soon you'll get the certificates.

[02:09:02]Day three attendance Q is already shared. In fact, it's already shared.

[02:09:07]Just scan it already. It is available on the screen.

[02:09:14]Yes, it is sharing now.

[02:09:54]I'll check I'll share again.

[02:09:58]I'll share again. H yes I have shared again. Just check.

[02:10:24]Okay. Those who have given your attendance uh you may leave. If you if you don't have any queries to you may leave You don't have any quiries and you are done with your attendance. You may leave.

[02:11:40]You joined your WhatsApp group. you will get updates regarding the uh next next series of workshops and next next workshop information.

[02:11:49]Make sure that you have joined the WhatsApp thereby you'll get the updates regarding the future workshops.

[02:12:52]Yes, hand back.