The Brothers Betting Their Robots Can Solve America's Welding Crisis | Path Robotics

[00:00:00]Welding is everything. So in 10ish years, we're going to lose 43% of our existing welding workforce.

[00:00:08]>> We want our systems to be able to weld at a superhuman level out of the box.

[00:00:12]>> I would say we were ahead of the current wave of physical intelligence, but we're actively also reinventing ourselves to be even better in that realm.

[00:00:20]>> Path Robotics is physical AI for manufacturing.

[00:00:30]We are already short according to AWS.

[00:00:33]360,000 welders for open positions. Just look outside. [music] All of the utility poles that you see that carry power to your homes, to your kids' schools, every building in the country all have utility poles carrying power. Those are all welded. My husband's a welder. I watched him go through a period [music] of tremendous growth in the business and get stalled.

[00:00:54]He absolutely hit a wall because he could not find welders and skilled labor and that was [music] the biggest bottleneck to him continuing to grow his company.

[00:01:04]>> Dad was an engineer and we grew up in the garage working in on machining parts and building little go-karts and dune buggies. So for us the go right to mechanical engineering. Starting there as a foundation made a lot of sense.

[00:01:18]Alice and I were doing our PhDs, talked to 100 US manufacturers, understand the biggest pain point, biggest pain point was welding. Then we had to go actually find somebody to pay us and then we actually had to go build this product.

[00:01:29]>> Alex and Andy came to my shop and they convinced me that if I gave them enough money, they would build me these two autonomous weld cells that would see, understand, and weld. And so it took us a little while. We signed a deal. Then about 6 years ago, they convinced me to come be part of this. Turns out that finding human welders isn't just a Joe problem. It is an everybody problem.

[00:01:51]>> Joe is customer number one. At that time he was president of course of manufacturing. He gave us $300,000 which at that point you and I were like 300 grand. We're going to be able to like build everything we possibly want. But then you buy your first two set of robots and you're like well that's $200,000 down to 100 grand. You know things get a little tight.

[00:02:11]>> And we immediately ran into a lot of hurdles. It was a hard time of [music] life to say the least. And so we just became very close. We were always close, but that really solidified our friendship, our brotherhood, like for life. And once you're in the trenches with somebody for 4 years at a young age, you can be in the trenches with them for forever basically.

[00:02:32]>> We'd have to actually see the thing we were going to be welding. And everything we were going to be welding was this very shiny stainless steel. We were just going to take like an off-the-shelf sensor, time of flight sensor, and just be able to 3D scan it. And that's what we would use to ingest that data into the actual robot itself and then say this is where we want to go weld and how we want to go weld. That was totally wrong. So we had to then on top of all the work we had to do build a brand new 3D sensor. So we had to solve this problem that no one had solved and we had about 3 months to do it which was [music] how do you 3D scan mirror surfaces. We immediately started like racking our brains on how we could possibly solve it. And the solution we came up with was let's make a 3D sensor where we project an asymmetric pattern and then if it starts reflecting on itself we can determine like what's the original signal versus what's like a secondary or third order reflection back to the system. That is what I would say is like the definition of starting a startup is like you go in head first [music] and you basically hit a wall on day one. And the way to survive and the way to keep going is just to be like think creatively, be super persistent, be adaptable, willing to like move and roll with the punches. But the big thing from this story is not just what it's like to be a startup. It was actually started the foundational layer for path.

[00:03:44]Day one, we were never going to make a sensor cuz we were like, "This is off the shelf. We can just buy stuff." But then this very first use case came with the first customer. And when you see it like that, when you actually interact with customers and see their pain points, you start to realize the edge cases that you never thought about in the lab. And that's what really kicked off the sensors are going to be fundamental to this company [music] from day one to forever. Because if this is the edge case that we're seeing now, what are the other 100 edge cases we're about to see with sensing alone when we actually deploy this at scale?

[00:04:16]>> So this is RO. This is our first mobile welding platform. So unlike traditional welding facilities and welding cells where you have to prepare the parts for the cell, this one it walks and approach the part themselves, it can even get inside a part to do the welding and that is a very unique and sometimes very necessary capability that our customers will require. It does not get tripped.

[00:04:36]You know, it's very stable platform.

[00:04:46]>> [music] >> Welding is actually pretty difficult.

[00:04:54]Imagine that you have a bowl of water.

[00:04:57]Replace water with molten metal. It's going to be, I don't know, 10,000° F and it's going to be brighter than sunlight.

[00:05:03]Take the bowl away. We have to shape this molten metal into its final position. And we have to make sure that it cools down with the same temperature that we want it to be. And it has to look good at the same time. It has to piece together two pieces of metal and you know hoping that they stay together for I don't know 20 years. So that's what obsidian is. Obsidian actually controls the puddle. It predicts what's about to happen next.

[00:05:29]>> For welding, you get about half a millimeter of inaccuracy [music] allowed and that's about it before the welds start failing or the welds are not perfect. The shape isn't right. the deposition doesn't go in the right spot.

[00:05:39]You don't get penetration of the joint.

[00:05:41]Extremely complicated process. That's why human beings still have to be in the loop. Not only do we care about the [music] geometry you're welding, we also care about the shape of it. We care about the angles and positions relative to gravity. We care about temperature.

[00:05:54]We care about the way the surface is prepped [music] because the surface itself can have huge influence on the weld. The space you have to handle all things you have to worry about is actually quite large and [music] it's not easily simulatable. That's why it's so important to collect data to build that neural simulator so they can then create a policy on [music] top of it.

[00:06:12]>> So for eight years, every robot that's out in the field has just been learning.

[00:06:17]And now we had this huge [music] foundation model that's built on millions and millions of inches of realworld data that we [music] get to play with and to use to train the robot how to see and how to understand what they're seeing and then how to weld and to learn like a human would learn. Our [music] customers range and are really primarily focused on energy and infrastructure on AI [music] data center buildouts and on ship building for defense. If we want to reindustrialize and we want to rebuild manufacturing [music] in the United States, the only way to do that is to create these real [music] world solutions that address the labor constraints that we have.

[00:06:56]>> But the way we do that isn't by going back to what we did once before. We are going forward in the future by using autonomous systems to make things. Human beings should be doing the creation part, thinking through how to make these things and [music] we should let the robot do the actual manufacturing process. And that's what PATH is doing.

[00:07:14]That's our mission. So we specifically focus on building a large model called Obsidian delivering our sensor [music] stack that gets all the input data from vision, sound, etc. that feeds that model that then runs and drives these robots with the end goal of giving our customers an outcome [music] that matters.

[00:07:30]>> And when I first met them, I thought welding was a sore problem. Using robots to weld that sounds like, you know, [music] 1990 something. But then I later realized welding with robots can only apply on, you know, a few applications where the parts are made perfectly or the fit up is so nice and it's so repeatable. It's not the case for the majority of the people that do manufacturing. It's such a huge unsolved problem. At the end of the day, the policy, which is a big network, says, "This is how I want to weld. This is where I want to go." And it just directs the hardware in that method in that direction. So, it can actually produce these fantastic welds out of the get- go. We're extensible. I mean, we're not tied to any particular hardware. It could be a a small arm. It could be a big arm. It could be multiple arms. Turn them into superhuman welders because again, it's just the [music] brain and the sensor.

[00:08:16]>> So, we have weld school where all of our employees, all of our new hires have the opportunity. We don't force anyone, but everyone has the opportunity [music] to learn from one of our very talented welding engineers quickly how to weld and then get to weld themselves. I'm horrible at it, but you all did an amazing job. I was very impressed. All right, S3, this is for you.

[00:08:42][music] >> Yeah.

[00:08:50]When I grew up, manufacturing was on a decline. We live in Ohio. We're very sensitive to that. I want that back.

[00:08:57]There's no reason we have to be dependent on other places around the world. We can do it here. I've got a lot of customers in the Midwest. And so, we decided to say in Ohio, in the Heartland, because we wanted to be obsessed with our customers painpoint, what really is their issue? And that needs to be a drive away, not a 5h hour flight away. The other thing is like this is shocking, but there's smart people everywhere. I know can't be possible there smart people in the Midwest. Oh, but there are. You know, we're 20 minutes from Ohio State. We're an hour away from Case [music] Western.

[00:09:31]We've got a ton of universities, Michigan, Georgia Tech, Virginia Tech, all within distance. And you don't have to be on a coast to want to solve hard problems, but you do have to be in the heartland if you want to learn about manufacturing. I almost don't even want to get the secret out about Columbus cuz I truly [music] feel like it's a hidden gem. I'm from New York and definitely one of my concerns with moving here was are there going to be cool companies?

[00:09:55][music] The reality is there are. So you get to work with incredibly smart people doing awesome things.

[00:10:00]>> The speed that we can operate with is a huge competitive advantage for us. And right now with all of the growth, especially in the data center and energy infrastructure space, which are two key verticals for us, we're able to move very quickly and meet that capacity need that these manufacturers [music] have.

[00:10:15]>> This is solving a real real problem today. So how do we scale that so that it is able to help [music] more and more companies? Where I see that in 10 years or more is we have a revitalized industrial base. We now build things again in the US. We're not as dependent on other countries. We're [music] not as dependent on supply chain weaknesses that we see. In December, I was asked to come and speak before [music] Congress.

[00:10:43]There's this thing called the Congressional Robotics Caucus. The caucus [music] is trying to figure out what are the areas that we should focus on to try and help this competition [music] with China. Like there's recognition that the government needs to do to come together with business [music] and figure out ways to help our companies accelerate adoption of robotics and [music] AI. But there's a lot of concern around AI. And so they asked four industry leaders to come and [music] speak before the congressional panel. And what a privilege to be able [music] to be in a time and in a spot where we're able to have those conversations and actually move the needle. You're talking about a stage with [music] Boston Dynamics and Google and PATH Robotics. So that is that is where we are and that's where we're going to continue to be at the forefront.

[00:11:35]>> The best case scenario when you guys come back it won't be one facility, it'll be multiple facilities. We're out growing our 200,000 foot facility here and we're taking the warehouse next door. Hope you guys see when you return [music] would be systems that are mobile. So, we're mobiley welding and and we have other systems that are doing things beyond welding. So, you will certainly see assembly. You'll certainly see other things in the fabrication process done either with a robot that sits on a track or done with a robot that moves around on legs. Best case scenario is 20, 30 years from now, US [music] is the the manufacturing powerhouse of the world.

[00:12:12]>> ALT Robotics was an awesome episode to feature on S3 a truly unique company in that they're in the Midwest, not just Silicon Valley or else Gondor or Texas, but in a place where there aren't a lot of other startups. I didn't get to go and do it, but what was it like filming it?

[00:12:24]>> Pretty epic. We went in planning to interview three, four people, the the founders, and we had like a line of people there wanting to speak with us.

[00:12:32]Their reasoning for being based in Columbus, Ohio was really [music] unique. We asked them, "Why aren't you guys in the Bay Area? That's where most of these venture capital firms are.

[00:12:38]>> They're an AI company, too. That's a big part of it, right?

[00:12:40]>> Why aren't they in the Bay Area?" And they said, "We want to be where our customers are." And they are, no joke, like within like a 50-mi radius of every single one of the customers that they possibly could ever want.

[00:12:50]>> I think a lot of what Silicon Valley [music] like AI approach gets wrong sometimes is like a lack of understanding of how hard the atomic physical real world [music] is. Yes, perhaps AGI comes from a lab in San Francisco and it just works with everything, but from our experience filming and working with like real physical stuff. I don't think it'll be that easy and you're going to want to do whatever [music] intelligence engineering and AI engineering you're doing in the loop of actual machines, actual robots doing specific tasks like welding. It is unclear to me at Path Robotics AI is better at welding than you two are. You guys got to weld while shooting this, right? How'd that go?

[00:13:24]>> We did.

[00:13:24]>> So, this this is what you guys welded.

[00:13:26]Did you both Is this who did what?

[00:13:28]That's was mine.

[00:13:29]>> I love you, Josiah, but um it needs help.

[00:13:32]>> I It needs It needs some help. But that's that's why there's welding school. That's why there's AI welding robots. Exactly.

[00:13:36]>> Now, you guys are too humble. But I did hear from them that the New York Times came in to [music] do a story on Path and they also welded. And they said that you guys were better welders. So, >> they did. We did. Yeah. Which [music] was a very nice accomplishment.

[00:13:49]>> Who do you want telling stories about your company? People who can't weld or people who can kind of weld.

[00:13:53]>> Can kind of. [laughter] All that put together made Path such an awesome feature on S3. Hi D. You you brought a box today.

[00:14:01]>> Hi, I'm Will, head of finance here at Story. S3 is an incredible but expensive show to make. And that's why today's episode is brought to you by Sen Cuts.

[00:14:08]>> What's amazing about Senkuten is they take raw metal and sheets or billets.

[00:14:11]They mill it, they cut it, and then they ship it to you days after you put your order in. It's fast, affordable, and made entirely in [music] America. So, if you're a team building some crazy project and you need metal parts, high quality and fast, reach out to Sanken and tell them S3 sent you. Don't eat the metal, Jason. Tastes good.