Episode Transcript
[00:00:00] Speaker A: I think it's becoming more abundantly clear to the world that when it comes to embodied AI, the winners in the space are going to be the ones who have the most real world data.
[00:00:16] Speaker B: Hello, everyone, and welcome to the robot industry podcast. We're glad you're here, and thank you for subscribing. Before we get to our episode today, I wanted to let you know we have a new podcast that I'm starting called Automation Matters. It's about the front end of the business, whether you're a builder, an integrator, distributor, or robot. Oem, this is for you. This new podcast is about sales, marketing, business development, strategy, and much, much more. And I'm excited about automation matters, but more on that later. My guest today is Rajat Bhagiria, and he is founder and CEO. And there's a funny coincidence that happened a couple of weeks ago. Three people mentioned chef robotics to me in the same week. And then I met Rajat at Automate Chicago. So this is not totally out of the ordinary. I was creating a webinar on food and automation and writing an article on the same topic. And today we're going to talk a lot about AI, about data, about machine learning, food end of our tools, and food is complicated. So, Rajat, welcome to the podcast.
[00:01:16] Speaker A: Thanks for having me, Jim. I'm very excited.
[00:01:18] Speaker B: So I made a mistake. I actually thought that you were not a roboticist, and you actually are a roboticist. So can you tell the audience a little bit about you?
[00:01:26] Speaker A: Yeah. Growing up, I always was kind of fascinated by, like, inventors. So the people who kind of, like, really inspired me were people like Thomas Edison and the Wright brothers and Ford, and of course, more contemporary versions of that, like Steve Jobs and Elon Musk. So I think I always kind of wanted to combine software plus hardware plus business, right? I think all three of those were quite fascinating to me. I think that kind of led to college and grad school. I went to Penn, and I kind of studied combination of computer science and I economics for undergrad. You know, I think that was really exciting. And during that time, Alexnet and Imagenet and deep learning was kind of starting to really become, like, more potent and powerful. So at the time, I actually started a small company in computer vision called Third Eye. And that was really exciting because, like, suddenly computers can make sense of the physical world. We were using computer vision to essentially recognize what's in front of you for the visually impaired. And this was a really good experience because it kind of got my feet wet in terms of computer vision. We had a small exit for this. And at that point, I became very convinced that perhaps the most powerful impact of software and AI and computer vision wouldn't just be in the purely digital world, but really on the physical world. You know, it's pretty obvious, right? But if you think about it, like, 90% of GDP is in the physical world, and probably the biggest part of that is labor, right? It's like 50 trillion of GDP in the world. So I think that very naturally led to, like, robotics. And that's kind of where I did my. A master's degree kind of, in robotics and machine learning. And of course, from that, chef kind of sprung.
[00:02:52] Speaker B: Oh, and that's a good segue, because my next question was about. Tell us about chef robotics and how you had this idea.
[00:02:58] Speaker A: Yeah. So I think the way this all kind of started is I became convinced that I want to do something about AI in the physical world, right? Embodied AI. And probably the biggest part of that is the labor market. And the next obvious question is, what's the right task to kind of focus on? So what I did is, I said, okay, well, what's the biggest opportunity? Right? I came from a technology background, not so much from the industry, right? So I was like, okay, where can I apply this technology? Right? So I kind of went to the bureau for labor statistics, and I was like, what is the biggest market for robotics and AI today? Which is basically, what's the biggest market in the labor force? And the biggest one was retail salespeople. And of course, that's a very human job. I didn't think it could be automated anytime soon. The next one was kind of nursing aids, personal aids, things like that. Same. Same story, right? It's really hard to do that. And then, of course, it was food service and food preparation. So from my perspective, it seemed like that was more tractable, right? Obviously, we all go to, like, you know, fast casuals, like Chipotle and things like that. And I was like, it seemed like we could do something there, right? So I started to kind of just talk to people in the food industry, like, what's the problem you're facing? And this was a very broad, broad exploration, by the way. This was everything from, like, food truck operators to fast casuals to ghost kitchens to, of course, food manufacturing. And I kept on hearing the same story over and over again, which is that there's this really, really big, crushing labor shortage. Now, obviously, like many of you are probably familiar, and, Jim, you're very familiar about this, too. Like, you hear this a lot in robotics. What I, what I learned in the food industry, though, is that it's a little bit worse than perhaps other industries like logistics and warehousing, because in food, it's either a very cold environment, like refrigerator. Refrigerator, or it's a very hot environment. So what I kept on hearing from everybody is like, there's a really big pain point around this labor shortage. Young people don't want to work these jobs. Baby boomers are getting older people, you know, that population is not going to do that. You know, there's a shrinking labor force participation rate, people are having less children, etc. Etc. All the trends. And because of that, what we're seeing is that people, we have big problems with kind of the supply for labor to meet demand. In other words, we're under producing. We could be making more, but we are leaving revenue on the table.
So I was like, ah, that seems like, you know, a great opportunity for AI robots, right? And by the way, this is across the board. This wasn't just like food trucks or fast casuals or food manufacturing across the board, right? And so then I was like, okay, well, is there a task that like, I could help with? And of course, what I learned is that in kitchens, basically everything except fine dining, there's like kind of three biggest kind of tasks that on a high level, right, 30,000 foot level people do. They do prepping of food, which is kind of washing things and cutting things. They do cooking of food, and then they do like plating of food. And what these operators in the food industry told me across the kind of sectors, right, is that the prepping part, while it's expensive, it kind of scales sub linearly. In other words, there's a lot of like more traditional automation, food processing equipment, things like that, that you can use. So like a couple people can kind of process the food for the day. You know, even if it's high volume manufacturing, it's the same idea. Of course the machines are bigger, but the same idea is true.
And then in cooking, same analogy is true, which is like, you can have, you can cook in batches, right? So one person can cook food for 50 people, 100 people, and manufacturing 1000 people. So cooking also kind of scaled sub linearly. So in other words, if you need more, more output or more throughput, you actually don't need more people linearly. It's more sub linearly. Yet when it comes to the assembly portion or the plating portion, that's really where everyone said, hey, there's a really big opportunity, and that scales more linearly. In other words, if you need more output or throughput, you need more people linearly scaling. So what I learned is that basically for everything except fine dining, if you were to solve any problem, really the right problem to solve is the plating or the assembly portion, which is really this task of given a pre cut, pre cooked batch of ingredients in, like, let's say, a hotel pan, your job is to kind of pick that ingredient, not damage it, kind of be consistent.
You know, place it into the right compartment. And, of course, you got to do that for hundreds of thousands of ingredients. Right. It's got to be flexible. So broad strokes, kind of like that exploration is kind of what led us to say, okay, well, let's find chef, and let's really focus it on kind of making this more generalized AI system around. How do you manipulate and assemble meals? And we could talk more about, like, kind of what customers are going after in the market were going after. But that's kind of how this all started.
[00:07:25] Speaker B: So what are the questions? And that's perfect analogy of where you're at. And I really love the fact that you've done your homework. Right. You've talked to a lot of people.
I wanted to ask you a little bit about your team and find out where you're located as well.
[00:07:38] Speaker A: Yeah, so we're in San Francisco Bay area. The reason we kind of picked that area is like, it's kind of an epicenter of, like, AI and robotics. From the AI perspective, there's a ton of really, really smart AI engineers from self driven car companies, predominantly, actually. And then, of course, increasingly now, there's a lot of companies that are using large language models, and San Francisco is really the presenter of that work. So there's a lot of good AI folks, and then there's also really good robotics folks. So for robotics, of course, you have a lot of different industries here. So you do have software in cars, which, of course, they have motion planning engineers. You have surgical robotics, which is a really big industry. Right. And a lot of those companies are in the South Bay. So we kind of picked this as our place to be. The other place we're kind of considering is Boston, which Boston has a lot of different robotics in these as well. The way I think about it is like, if you were making art in the Renaissance, where would you want to be?
[00:08:35] Speaker B: Right.
[00:08:35] Speaker A: You would want to be in florence. And it's like, if you're doing like, a. And robotics in 2024, where would you want to be? It's like Bay Area. Right.
And by the way that's very kind of cleanly, kind of kind of segues to our team. So if you look at our team, it's like 95% computer vision robotics software engineers from top AI robotics companies. We are trying to take more of a software driven approach. The team is mostly a software team, and, of course, there needs to be harder people as well. One of the things we've done is we've gone to a lot of self driven car engineers, and we said, look, obviously, autonomous vehicles are going to be huge. We all believe that. But it might take ten years to get there. Why don't you come to chef, and we're going to give you a problem that we can take the exact same skills, your exact same computer vision and perception and motion planning and path planning skills, and let's apply it to a problem where there's a market today, it's a hair on fire problem. It's a big market. And by the way, we can actually ship robots today or soon, right? Not. Not in ten years. So that's kind of how we think about our team.
[00:09:31] Speaker B: So we talked in the warm up that you've been in stealth mode, and, I mean, a lot of us understand why in stealth, but why stealth and food?
[00:09:41] Speaker A: I think the reason for this is in food robotics. I think a lot of folks kind of start with I want a cool technology, and they're very much like very technical founders, and they kind of. They end up kind of building something cool. But that doesn't necessarily solve the right problem. What I mean by that, and there's many people who have kind of built robotics for restaurants, for example, the volumes are not crazy high. People always talk about, like, specialized roles versus generalized roles. So, in manufacturing, right roles are more specialized, like Sally's on the conveyor belt, and Sally's picking at station one. And she's doing this task, it's probably going to be one ingredient. Whereas kind of in a restaurant, right, what you find is that because the volumes are a lot lower, the rules are a lot more generalized. So now Sally's not just doing one ingredient. She's walking your container, your tray, your burrito. Down the line, she's doing 100 ingredients, 50 ingredients. So the autonomy level needed to get to 100 ingredients or 50 ingredients is extremely high. It's a very hard technical problem. And by the way, to be fair, Sally's probably also cutting ingredients. She's also cleaning the floor. She's also cleaning the bathrooms. If you want to have a Sally equivalent, which is what you need to get an ROI. The autonomy task is very high, arguably insurmountable today.
You basically need AGI to make something that's Sally equivalent in a restaurant because she does so many tasks. Right, right. And I think we were seeing that, and we're like, look like we don't think that's the right approach. And, you know, really, like, that's. That's, of course, the vision, by the way, the vision is really a robot in every commercial kitchen. But we thought that everyone's trying to build the model three. And really, what the industry reads needs right now is a roadster. A roadster where you can get to market. This is like a test analogy, right? Like a roadster, you can kind of get to market, add value to customers, and of course, by adding value to customers, you can scale more robots. And by scaling more robots, you can create this data engine, which really allows you to do the harder thing. And so I think we had taken that route. We found a really good market, which is food manufacturing, that we thought that not many people are thinking about. And we wanted to keep that under the radar for a while.
I think what changed, and the reason we're now more open about what we're trying to be more open about what we do, is, I think it's becoming more abundantly clear to the world that when it comes to embodied AI and real world AI, the winners in the space are going to be the ones who have the most real world data. And what I mean by that is, if you think about a company like OpenAI or anthropic or one of these large language model companies, the crux of a lot of what allowed. A lot of what allowed these models to really exist is like hoards of data, right? Absurd amounts of data. And for those companies, it's just, you download the Internet, and that works fine for other companies. If I think, like, I think if I think about a company like Aurora or Waymo or cruise, a lot of that data can be garnered in simulation, right? A lot of their training is in simulation. Now, of course, they get some real world data, too, but a lot of it's in simulation. And for yet other companies, they can really. They can really, like, get a lot of the data in the lab. They can kind of, like, have a test set up, and they can kind of generate the data in a lab. But I think when it comes to food robotics, I think what you learn is that none of those really work, because simulation is not good enough for deformable food. And we saw that, and we're like, oh, wow, we have something really cool to offer here, which is we have deployed dozens of robots all over North America. We made now almost, I think, more than 17 million servings in production. So we feel like we have an interesting story that can be exciting for other founders, other roboticists to really do. And that's kind of like the thinking of going more public is like, hey, look, to succeed in this space, you need real world data. The way to get this is just deploying lots of robots. And here's a kind of cool example of chef that's kind of taking that path.
[00:13:27] Speaker B: So I have a question about why is AI so critical to the mission? And you now have served 17 million meals. So are you kind of the leaders in food automation, AI right now? And as you scale?
[00:13:42] Speaker A: Yeah, it's a really good question. So, by the way, that was exactly the question I had when I started this company. Right. Which is like, like the obvious question to ask is, why hasn't anyone done this, right. It seems so abundantly obvious. And what I learned is actually that there's a lot of traditional automation that does exist in the food industry. And I'm, I'm sure you have watched some of these how it works videos of these depositors and dispensers, right? There's, there's a lot of systems, right? If you were to go to something like a pack expo, right? Like, you see lots of traditional automation, right? You see laboratory feeds and piston based dispensers and volumetric dispensers and augers and all these kind of systems. But when you talk to those vendors, what you learn very quickly is that the systems are very, they really work well for low mix, right? In other words, like, if you want to do the same thing over and over again, great. Like, for example, if I'm Kraft Heinz and I'm making bottles of ketchup, then of course, there's not that, you know, within that skew, I have extremely high volume and I have relatively low product mix. I'm going to get a custom machine, a custom line from systems integrator, and it's going to just do that, Heinz ketchup all day long, 24/7 days a week, right? So that's what kind of exists in the market today. But what we find when it comes to meals, right. Whether those meals are like a sandwich or a salad or a prepared meal or a wrap, what have you, right? Consumers want choice. Consumers want veggie meals and they want meat lovers meals and they want gluten free meals. And of course, those meals change every season, and sometimes they change every year. And the point is that if you look at the meal market, there are companies with hundreds of skus. And if you think about something like a ghost kitchen, I mean, every single person wants a slightly custom chipotle bowl or a sweet cream bowl. So I think the point is, when you have all that mix, traditional automation is not flexible enough. And I think what AI allows you to do is AI allows you to take more traditional hardware, like a robot arm, cameras, compute, GPU's from Nvidia, and really make a system. And now that system, you can kind of productize, the harder you can productize and mass manufacture. And then what AI allows you to do is deal with the variations in food. The thing with food is it's organic. So if you take an onion and you julienne it versus chopping it into little bits, it has very different dynamics. It's a very different material. Now, if you saute that onion for a couple minutes, a siphon and saute for ten minutes, if you broil that onion, it's a little bit different. Now, if you put it in the fridge, it's even more different. And by the way, Sally cuts it differently than Bob and Bill in the Idaho plant is a little bit different than the way that Lanta plant works, etcetera. So there's just so much variety. And what AI allows you to do is sense the world. It has cameras, and you can sense world. You can make decisions intelligently using these new deep learning and transformer architectures. And you can actuate. So you have more intelligence, you can react to the differences. And I think that's really what allows this kind of high mix application that usually wasn't possible to be done, to be possible now. And, you know, within that, yes. At least as far as we know, we have, we believe we're the market leaders. We've done more mills in production than, I believe, arguably all the other companies combined.
[00:16:50] Speaker B: I'll say everybody else.
[00:16:51] Speaker A: Yes. Yeah.
[00:16:53] Speaker B: So the fact that you have got this AI big vault that actually helps your customers. Right. So customer a and customer b, you guys are sharing data a little bit, because it's about food. There's nothing competitive about corn or beets or celery or whatever.
[00:17:10] Speaker A: That's right. Yeah, exactly. And of course, just to be very clear, it's not like we're taking any proprietary data, like recipes or anything. Like, we have cameras that, like, literally just looking at, like, your bed of shredded carrots. Right. So it's really kind of like we take that kind of data about how the ingredient looks, the material properties of the ingredient. And of course, as the robot plays with it, we have lots of different sensors that kind of learn properties about how the ingredient and makes this core AI brain. We call that core AI brain chef os better. And yes, the idea is that the more these robots we deploy, the more that every one of our customers benefits. The AI gets much better. It's able to react to more edge cases. It's able to react to like, oh, today the cook added a little bit extra oil, so the rice is a little bit less sticky. Right? Or today the sauce is a little bit more watery. We can react to that. Whereas usually traditional automation systems would not be able to react. Of course, humans can very trivially react because, of course, they have these very intelligent brains, but it allows really kind of automation to react to that.
[00:18:11] Speaker B: And you're keeping. If I'm one of your customers, I'm not. You're not touching my recipes or my process. It's just really about the data, about the food and the shape and the texture and all this kind of thing, right?
[00:18:21] Speaker A: That's right. That's right. Exactly. And, like, that's right. So, like, you know, we give our customers basically, it's kind of like we almost think of ourselves as like a robot staffing agency in some regards. Many of our customers do use staffing agencies. Just given that it's really hard to hire even there. It can be hard to hire even with those agencies. There's a lot of turnover and stuff. So, you know, we say, look, guys, like, we're going to give you this kind of, like, turnkey system. You can just slide it onto your line and you'll tell it what ingredient you want it to run, what portion size you want it to run. You'll tell it what tray you want to place into, what compartment you want to place into. You press play, you load the ingredients, and that's about it. And then when you want to run something else, it's equally simple. You. You say, okay, well, now I'm going to do this other ingredient. You know, you plug in the two hotel pounds of food and you run that.
[00:19:04] Speaker B: Rajat, one thing we didn't talk about is about how horrible the labor shortage is in at least North America.
[00:19:12] Speaker A: It really is, right? So, like, you know, many of our customers today, they run in cold rooms, right? So, like, it's anywhere from 34 degrees to, like, 38 degrees fahrenheit in these cold rooms. And, you know, it's really not great working conditions, right? Because you and by the way, there's no alternative, right? So, like, it's not like, it's not like there's any alternative, but, like, the reality right now is that you have to go into these cold rooms and you're kind of doing these scooping motions for in eight to ten hour shifts. And usually there's two shifts a day. So, you know, it's really hard to hire for these roles. And even if you do hire, I mean, oftentimes people just, like, leave after a day.
[00:19:47] Speaker B: They just quit, right?
[00:19:48] Speaker A: Yeah, they just quit. I mean, there's literally multi hundred percent turnover rates. So, you know, what we often find at our customer sites is we'll go to a plant and, you know, like, they'll have x lines, right? Whatever x might be, depending on the volume. And they're only using 60% of x, right? Like, the other lines are just sitting stationary. That's usually not because they don't have demand. They have plenty of demand. It's just that it's so hard to hire, they can't run it. And by the way, it's kind of like if a line needs 20 people to run, but you only have 15 people, you can run, but now those people are doing a ton of overtime, and then they're going to quit because of that. There's all these issues basically because of hiring. And I think what we have found is that our ROI for our partners and our customers is not so much about cost savings. Obviously, we do save the money. Yes, we also save them on yield. There's all these other things, but really, more than not, we really help them increase volume by just running more. They can just, instead of running 60% of the lines, they can run 100% of their lines. Instead of running one shift, we can help them run two shifts. Instead of running only one plant, we can help them open up another plant, because instead of hiring 300 humans, which is hard, hiring 100 robot workers is much easier.
[00:20:59] Speaker B: So let's talk a little bit about the robots and the automation. What does an automation system line look like from chef robotics?
[00:21:08] Speaker A: And maybe before that, I can tell you about status quo. So if you were to go to one of our customers, you'd see long conveyor line. There's usually either an automatic de nesting system for tray denesting onto the conveyor, or there's a human who is placing the containers onto the conveyor. Then each person basically has one ingredient, and they're scooping the ingredient from a source container into the customer tray that's moving down the line and then there's a refill runner, and these refill runner. That's a separate kind of role. And these refill runners are kind of refilling source containers with the human depositors, basically. So that's kind of status quo. So what we essentially make is a chef module. Our modules are essentially the same footprint as a human. And of course, that's very much by design. And the way it kind of works is that you say, hey, Bob, I have a different task for the plant. Something more high value, or we think we can leverage your skills, and we're going to have the robot do that task for today. And you basically slide the robot to the line. It's the same footprint again. Maintenance team or the production team plugs in 110 ac and then a compressed airline. That's kind of it. Those are the only two inputs we need. And like I said, you load the ingredients and you press play and then shuffle with the pause. That ingredient. The reason I start is because what I just alluded to here is you can start with as little as one robot. And by the way, I think that's kind of interesting, because in traditional automation, it's much more common to fully automate the line, the process. Sure, I think there's a lot of pros with that, by the way, but there's a lot of cons with that, too. And to name a few. Like, you know, from an autonomy perspective, if I have to automate the full process, I have to handle every single in the food world, every single ingredient. So let's say there's a really, really tough ingredient. Like maybe they want to spread, like, avocado really nicely. It's a very soft ingredient. They want to spread the avocado on the top of the tray. Obviously, that's just hard, right? Like, humans have very soft fingers, and that's where they're able to do it. If we wanted to automate that line, we know it was a fully automated line. We couldn't do it. But now what we can do is say, hey, let's put five robots that do the starch, the protein, the diced vegetables, the sauce, the garnish, and maybe a human can put the avocado. So it allows a lot more flexibility.
The other thing that's really nice about this kind of partial line format is when it comes to downtime, if you have a fully automated line and the robot goes down, the line is down. I mean, you're losing potentially millions of dollars with chef. What's nice is that, let's just say for whatever reason the chef system on a particular day, on a particular line is having some troubles. Fine. And of course we make it very easy to troubleshoot. But let's just say that's not sufficient. You just pull the robot out and you put a person in. Obviously that's not great, but the point is that the line does not go down. So I think our model really is around this partial automation, which is, hey, let's deploy these robots. Let's start with a starter pack of six robots that's good enough to really prove our ROI, add value, and it's relatively low risk for you. And then of course, as we kind of prove ourselves, then you can very quickly scale up. And on our side, of course, the more of these robots we deploy, the better the autonomy gets. And the better the autonomy gets, the more utilization these robots are going to get, which means they're adding more ROI for our partners and the systems get better. And now it makes more sense to get more systems. So that's kind of how we think about that.
[00:24:32] Speaker B: So the more you use it, the better it gets. The more you'll use it, right?
[00:24:35] Speaker A: Yes, that's right. That's right. And then that flywheel, by the way, is really important. That data flywheel is like very critical for the AI, right? Like, I think, like, you know, what you'll find in the AI world is like, my belief at least is the models themselves will become commoditized. The hardware, yes, it's a little bit hard to access GPU's right now, but that's, you know, that's going to go away, right? There's really not a ton of moats except the data. Right. It really is just data. So if a company can create a data flywheel where the more these systems are deployed, the better the systems gets for our current customers, which means that those current customers buy more. That's really powerful.
[00:25:09] Speaker B: Thank you for that. Now one of the things you mentioned to me that I don't know you mentioned yet is the conveyor system is a little bit detached from the chef robotics system. You're the expert, you tell me about it.
[00:25:21] Speaker A: Yeah.
When it comes to our production customers, it's in food manufacturing. They already have a conveyor. The humans are depositing onto these conveyors. So what we basically do is we just slide onto the conveyor and there's a 3d camera, an RGB D camera, so it has depth right on top of the conveyor. So we're kind of detecting and tracking these containers. Now there's a bunch of different benefits to this, right. When it comes to the customer benefits, it's really nice because if they change their containers, which is not infrequent, by the way, we don't need to change a bunch of stuff. We retrained the computer vision detector and now 3 hours later, we're off into the races, running with their new container. Or maybe they have a new customer and that new customer has new containers. Fine, same deal. Sometimes we have customers with multiple kinds of conveyors throughout the plant. So one conveyor is kind of a dark blue, and then there's another light blue, and then maybe there's a white conveyor, there's a slotted conveyor, etcetera. We don't need to make custom hardware or custom software. We just have a computer vision model. You retain it, you're done. Then on our side, of course, what's also nice about this is this allows us to scale across customers, too. For us then to go to a different customer or different plant, same knowledge. We don't need to make custom software and stuff. We really like this software based approach because it's a lot more scalable. It also deals with edge cases. We have a customer wear photo eyes, basically for detecting when a container comes. But it's very fragile and like, you know, sometimes a piece of food or some sauce spills, you know, right there. And then the photo is just kind of going and spilling sauce everywhere or ingredient everywhere. Whereas with shaft like, because this kind of like aerial camera, you know, if the container is a little bit rotated in terms of orientation, it's a little bit farther from the conveyor. We deal with that, and, you know, if there's no container, we won't just place on the conveyor. So I think it's just a much more scalable and like, robust system. So that's kind of why we decided to go that route.
[00:27:14] Speaker B: No, it's great. It's really. Detaching yourself from the conveyor, I think makes your life easier at the end.
[00:27:19] Speaker A: Yes.
[00:27:20] Speaker B: Rajat, what are some of the questions? You must get a lot of questions from prospective customers. What are some of those questions?
[00:27:27] Speaker A: Like AI and robotics is something that many of our customers are not extraordinarily familiar with. They likely have thought about robots when it comes to secondary and tertiary packaging. When it comes to case packing and palletizing. Right. They probably thought about that because that's something that traditional automation can more can help with. But when it comes to primary packaging, they're usually not super familiar with AI and robots. So the very first question is usually, hey, I've never really seen anything like this. Does it even work? Right. It's kind of a disbelief. Right. And they're like, how is it different than dispenser? Which, of course, is the right question to ask, which is exactly what question we had as well. So I think a lot of the question that we get is, like, let's talk about, like, how you solve similar ingredients for other partners, case studies, things like that. So I think that's usually probably the biggest thing. We kind of talk about other things that our customers really do care about is food safety. I think that's something that really frequently comes up, and we can talk more about that here, too, but that's something that really, our customers also care about.
[00:28:23] Speaker B: So let's do that. Let's talk about food safety. And you obviously have to adhere to industry regulations. And probably, in a lot of cases, you guys are the experts, because maybe they're experts at making food, but you're making food with repetitive motion robotics.
[00:28:39] Speaker A: That's right. That's right.
And food safety is super critical. And I think we really tried to partner with our customers when it came to this stuff. In other words, we had some early design partners, and these design partners really helped us through this food safety process. And then, of course, as we deploy more robots, we hopefully gain even more expertise in them. As you alluded to, the way we think about food safety is because we kind of have a pick and place based system, right, as opposed to a dispensing based system. The only thing that's actually touching the food is really two components. Number one, the hotel pan, and number two are utensils. Of course, the hotel pans are NSF and off the shelf. And so that's kind of a done deal. And then when it comes to these utensils, we really try to design them with kind of hygienic design in mind. So, for example, you can, the utensils are toolless. You can really quickly remove them from the and effect it by just pulling out two pins. And you can give it to your sanitation team, and they can either dishwash it or put it into a cop or clean out a place tank, and off you go. So that's. I think that's probably the key of the whole thing, which is the only thing that's actually touching the food is two components, and they're easily removable. They're made of food safe materials like deloraine, the utensil is, and titanium. And they're also made of food safe. They're made of food safe manufacturing pots. So, for example, like, just to put that into perspective, like, if you were to 3d print a part that's not usually food safe, especially by, like, FDM or something. But even if you use a food safe material, if the manufacturing method is not food safe, then it doesn't matter. Whereas, like, what we do is things like wire, EDM, and CNC machining. And that's more of a food safe manufacturing process when it comes to food safe materials. And then for the rest of the system, of course, we design it. It's called, like, a splash zone. So there's a direct food contact, which is the utensil in the pen. And then there's, like, splash zone. And for the splash zone, I mean, it's really kind of picking the right materials and the right design. So the vast majority of her system is made out of stainless steel, so that really helps. And then outside of that, we really follow the guidelines. So, for example, rounded corners, no sharp edges, no sandwiches where bacteria can be harbored, things like this. So I think food safety is one of those ever changing things where we went through the ringer with some of our early partners. They really helped us make their initial system. And then every single partner we meet, like, they give us some more feedback, makes the system better. And then now we're generally finding that, like, we don't really get, like, we kind of talk them through the food safety story, how we think about it, and people are usually on board, which is great.
[00:31:02] Speaker B: So I love this question, by the way. I'm going to ask you next. What has there been a food that's given you a lot of grief? Like, is there some kind of food geometry, like broccoli or something, where it's like, oh, my God, our vision guys are, like, driven crazy by this food.
[00:31:18] Speaker A: Yeah, yeah. There's many. Honestly, the thing with food is it's so variant, and there's a lot of trouble that comes with that. So I'll give you a few examples. So broccoli is actually a really good one. Cauliflower, broccoli. I think there's a few things that are hard about it. So one thing that's hard about that ingredient is, like, you usually have a target weight and you have some tolerances. Right. The thing with broccoli is that since the particle sizes are really big, I mean, you get, like, one particle less or one particle more and.
Yeah, yeah, exactly. It's like, it's. It's almost like a. Like, it's kind of like, accountable. You have to count the number of pieces. And then, of course, another thing that's tough about broccoli and cauliflower, some pieces are really small, some people pieces are really big. So I think that's honestly just a tough ingredient. Another one that's been tougher than we expected is something like pulled meats, pulled pork, pulled chicken, things like that. One of the things that you realize when it comes to, like, meats is that the meat actually changes material properties even throughout pan of food. Maybe there's, like, multiple animal animals in that order, or the same animal actually has different densities throughout throughout it. So, like, you know, a pan of whole chicken, you have to dynamically change how you pick based on where you are in the pan, which is interesting.
[00:32:26] Speaker B: Okay.
[00:32:27] Speaker A: Yeah. And then, by the way, like, there's a more higher level thing, which is, like, one of our customers, a lot of cheese grits. And by the way, like, every single time we manipulate those cheese grits, they're a little bit different. And that's not because the customer wants them to be different. It's just because a human is cooking them. So there's that broad scope thing, which is like, humans make food and they're variable a little bit. Right?
[00:32:47] Speaker B: Yeah. And the people, like you say, they cook them differently, and so it changes everything. So can we talk about maybe change to spices? Well, we talked about this in our, in our pre call is what about spices and how do you deal with that?
[00:33:01] Speaker A: Yeah, so, you know, we were kind of surprised, actually. But, yeah. Like, even spices are usually deposited by hand. And by the way, there are dispensers and depositors that can do spices, but our customers still choose not to use them oftentimes because of the high mixed nature. In other words, they're not just doing oregano for 24 hours. They might be only running that meal for 6 hours, and then they'll change over and they'll do time. You get the point. So even though traditional automation could do it, they don't use it because they want to change over. And that thing's going to take 3 hours to clean. So the way we think about spices is we essentially have a very small utensil. It's like, literally half a gram. We can pick up, like, as little as half a gram of spice or garnish, really. And, you know, it has a small self cleaning mechanism to really, you know, prevent the spice or the buildup. Exactly. And then after that, it's kind of the same as really any other ingredient. Right. Like, we'll use computer vision to understand the topography. Once we understand the topography, each of our ingredients has a policy, a manipulation policy. And we will use this policy to figure out what pose or like point and orientation that end effector should look at in the hotel pan to be consistent. Then we'll plan a path that point. Once we plan a path, that point, we'll execute the pick. And we actually have scales underneath our hotel patents. And these scales actually allow us to close the loop. These scales allow us to say, oh, we thought based on the policy that we're going to pick x grams, but actually pick y grams. And that's a little bit off. So now let's adjust. We can either reject that pick if the customer wants us to, or we can now also change the policy, fine tune the policy for the next set of picks. So really it's actually like the biggest thing that's different from spices to other ingredients is number one, the policy and number two, the utensil. And by the way, that's kind of a general statement for chef. Like we can really work with any most, I would say, ingredients you want us to work with with different policies and utensils.
[00:34:50] Speaker B: Right? So if I'm a user of chef and I decide I'm going to make sticky rice, then I just add it to the system and you could tell us a little bit about how that works.
[00:35:00] Speaker A: Yeah, it's a good question. So the way it kind of works. So I can tell you about what it is now and I can tell you about what it will be very soon. So the way it kind of works right now is we give our users a web app. And on this web app you can kind of log in. And usually our customers like a meal card. And that meal card will say, hey, here's a meal, here's the SKU number, here's the ingredients, here's the portion sizes, here's the container quadrant or compartment in the container, etcetera. So they upload that PDF. And we're going to use a large language model. We're using a large language model to kind of parse that. And then what you can do is basically on a per ingredient basis, you can say, okay, well, for the sticky rice, and we already have a little bit information from you. So maybe we have information about how it looks from the meal card. Maybe we have a portion size. And of course we have the text, we have the string name. So from that we will use a VLM vision language model to generate a few images and basic questions to ask the user. So we'll give the user from edamame, which is not sticky, to cream cheese, which is extremely sticky, where does this ingredient lie? And we'll show you different images and you can drag the slider and we'll ask you three or four questions. And from that and the image and the string of food, the name of the food, we will basically generate a policy, initial starting policy, of how do you manipulate that? You'll do that for all your ingredients. It's actually a relatively quick process. And then you will deploy that to the system. That's the first part. The second part is basically the user will slide the ingredient, actually, into the pan, into the system. And the system will basically fine tune. It'll pick and dump, is what we call it. Pick and dump from one pan to the other, over and over and over and over and over and over is not a time of time. It's like a few minutes, by the way, and it'll basically fine tune. So it's like, hey, I was starting policy about a minute ago, the sticky rice. Based on my prior experience, when you're building sticky rice now, let me fine tune, because, of course, this particular sticky rice that was cooked on this particular day by this particular person and this particular plant is a little bit different. So let me fine tune that and then you can kind of run production. So that's kind of how it works today.
What we are building now is kind of actually leveraging more learning from demonstration and diffusion policies and transformers to basically have the user do some demonstrations. The user can actually demonstrate scooping sticky rice with a robot, and it'll learn there. So that's something that's really just meant to make it simpler. I think probably the reality is it's going to be a little bit of both of these. There's a lot of good that the current system works, does, but the second system is hopefully something that can even further make it easier for our users.
[00:37:32] Speaker B: Thank you for that. Let's talk a little bit about the actual robot now. And you say it's human scale, so it's not huge. And let's talk, too, about your controls. And of course, we come from a lot of PLC's, but you're not using PlC's anymore, right?
[00:37:47] Speaker A: That's right. So, yeah, we don't use any PLC's. Industrial PC we have in each of the computers and each of the robots. Yeah. The entire system is kind of running on Python and C code, you know, for. We are based like our. For our pub sub and our communications, we use ROS.
So ROS is really what helps us kind of like communicate to all the different nodes, whether that's like a camera capture node or kind of some of the compute nodes to actually make some of the calculations. So yeah, like RoS is kind of what helps us kind of do a lot of the motion planning and then the majority of the code is really written by us, right. To actually kind of do the perception and the controls and the motion planning and all of that work.
[00:38:27] Speaker B: And you've got customers because you've made like 17 million meals. Are you like telling people who they are or can you tell us a little bit about what kind of industries they're in?
[00:38:40] Speaker A: Yeah, so in terms of the industries that they're in, I think we found success in a few. So one is kind of frozen prepared meals. So I think that's one good example of customers we have a good amount of success in. Another one that we found good success in is fresh prepared meals. So the big difference of course is frozen, you know, goes into your freezer. Fresh is more like something you would find, for example at a hospital or maybe a school or a cafe or cafe, you know, a gas station. It's, it's more like a fresh salad or it might be like a wrap or a burrito, things like that. Right. Fresh also, by the way has things like sandwiches and party trays and yogurt parfaits and fruit trays and things like that. So that's kind of the fresh world. And then outside that, we've also found good success in contract manufacturing. So contract manufacturing would be, you know, I am a brand but I don't want to do my own production. So I'm going to partner with somebody and they're going to kind of do the production for them. And that, by the way is not totally dissimilar from kind of the frozen or fresh world, but there's a lot more variety. So now you have different trays.
Different customers care about different things. Like some customers really, really value consistency. Other customers really, really value throughput. Obviously everyone wants everything but they care a little bit more about different things. So we have to work with, it's like we have a main customer which is a CM and then they have their own customers and we kind of have to make everybody happy.
We also found some good success in direct to consumer. So these are companies that will prepare meals in a central facility and they'll kind of ship them to individual customers via FedEx or EPS. And then increasingly we're getting into kind of food service also kind of things like ghost kitchens. Well, that's kind of the ghost kitchen is kind of what's next.
[00:40:22] Speaker B: Right. So I was just going to ask who's on your neighbor and who's on your radar for bullseye type customers, but food kitchens, certainly ghost kitchens, that makes a lot of sense. So when you're. You don't sell these systems, right? You rent these systems. Can you talk a little bit about, about that?
[00:40:39] Speaker A: Yeah, for sure. So, yeah, so our model is a little bit different than many automation companies, which is we provide robotics as a service, which is to say we won't actually charge you any capex upfront.
The way it kind of works is you're going to pay a yearly fee for the chef's service. And that basically means that we're going to take care of everything end to end for you. And that, of course, includes the hardware, includes the software, it includes all the servicing. So for example, we have 24/7 support line. We're going to have local texts in case something goes wrong. But perhaps more importantly is we're going to provide very frequent, on the order of weekly software upgrades and also hardware upgrades. The big idea here is that to do high mix, you need software. Software is not a nice to have, it's a need to have.
The thing with training and inference in machine learning is it's expensive, it's costly, there's server costs, there's real server costs. So it's not like we can just ship a piece of hardware and then kind of forget about it, right? Like if you're changing your ingredients, you're changing your menu, which is, again, these are high mix customers.
You need to update the models. And the only way for us to do that is we got to pay ourselves. So that's kind of why the recurring makes sense. And then of course, what you get out of that as a partner, and the way we think about it at least, is that we're in charge of constantly helping you improve production. So giving you more ingredients, helping improve your consistency, helping improve your yield, helping improve your throughput, helping reduce billage, things like this. And oftentimes, by the way, that doesn't just mean software, it also sometimes means hardware. So as we're all very familiar with, there's this company called Nvidia and they're constantly making new GPU's. And as AI models get better, yeah, we can give you an AI model, but if the AI model can't do inferences well on the system, on the hardware system, the edge, right, then we also need a new GPU so, or for example, maybe to take advantage of some of these new models, we need better cameras, we need higher resolution cameras, or maybe we need new sensors to get better force torque sensor readings.
So the point is that the way we think about it is we're providing the service, which is we are going to make production work. And you, customer, are kind of, you kind of like, we are kind of your outsource robotics division, and we're going to give you software, new hardware, we're going to help you kind of manage the high mixed nature of. That's kind of how we think about it, honestly, it is different for the industry. It's a new, new thing for the industry. But I think the good news is that, like, now that we have kind of really established ourselves as like a company, that's like working with some big enterprise customers, like, I think we're getting a lot more buy in from our partners or why this makes sense. And they're kind of understanding, ah, this is like why this makes sense.
[00:43:24] Speaker B: So I'm kind of interested a little bit, and I think it's great. And I think your approach makes so much sense. Like, I, I've been in factory automation my whole life, and, you know, if the vision system didn't work in a factory, the guys would turn it off. Right. And so that just, it's just a road to, you know, where, um, when, you know, how does that, if I'm interested, what do I do? I call you guys up, I fill in a form. Like, what happens? Do I do a visit?
[00:43:49] Speaker A: Yeah, exactly. So usually you'll get in touch with us, and at that point we'll have an intro call where we learn more about the application.
We really try to make sure that our thinking is we're not going to get into a partnership unless we really think we can help. So we'll really try to understand the application. Obviously, we'll tell the prospect about us. And at that point, the very next obvious step is, hey, let's sign an NDA and let's fly out there. That's like step two, let's just fly out there or drive out there. So then we'll kind of do a site visit, and we'll really get to see production. We'll really make sure that, hey, look, our technology can help, and we can actually add value. Assuming we can do that, then of course we say, okay, great, let's put together a proposal. And then we kind of, if there's any discussion that needs to happen there, we'll do that. Oftentimes our customers will send us some of their ingredients, and we'll kind of do some technical validation. And as soon as we do this technical validation, we'll sign an agreement. And once we sign the agreement, the customer will kind of put down a fee for the initial kind of the configuration and deployment. And that's not a super high fee, by the way. It's just meant to say, hey, look, there's going to be a bunch of, we're going to have to like, yes, I'm serious. And we're going to be living there, just deploying these robots for you for a bit of time. And then, yeah, we're kind of doing development throughout the entire phase. Right? We're doing development. We're configuring the systems, we're onboarding ingredients. We were testing the ingredients. And then essentially, once we feel really good, we'll create the robots up, we'll put them on a truck, and we'll kind of ship them to the customer site. As soon as they get to the customer site, you know, we'll fly out one of our applications. Engineers will fly out there, we'll uncreate the robots. And usually in food facilities, you have to, like, you wash the systems, you swab them to make sure there's no bacteria, hard bridge, and then you kind of move them to the line. And we'll probably spend like a couple of days, two to three days, kind of doing initial testing just to make sure there's nothing kind of like unknown, right? Unknown unknowns type of thing. Once we're kind of through that, we will do a site acceptance test. And that site acceptance test is like, hey, let's really prove to both of us that this thing works as per kind of the understanding you had. And then as soon as that is met, that's when the customer says, okay, I'm happy. They kind of pay that first year fee. And at that point, basically, we have trained their team, their production team, and their planning team on how to use the robots so they can run unsupervised is what we call it. So at that point, basically, our team will probably fly out. It's kind of like when things are stable, their team is running and it's stable. And at that point, basically what we do is we do a couple of things.
We have weekly meetings with them, everyone we're partners with weekly meetings where we talk about day to day kind of issues, any kind of day to day issues, and also more the success part of it. Hey, what are the more macro things that are not issues, but we want to get ahead of it. Our goal, of course, is like, hey, how do we kind of scale to more of their plants and more robots within their plants? And then, of course, their goal is, hey, how do I kind of get best advantage of these robots? And then, you know, we'll do things like 24/7 monitoring. Like we have automated alerts, so anytime something goes wrong, we will get a notification on our slack. And whoever's on call say, ah, there was a fault. Okay, let me, like, log on, and I can look at that timestamp. You know, the robots are sending data to the cloud, and we can say, okay, from like, 04:32 a.m. to 04:34 a.m. there's this fault. They can analyze it, and they can say, hey, it's a user error. Hey, customer, you should not do this. Or they can say, hey, there's some bug. Let me fix it. So that's kind of how it works, basically. And then, of course, if there's any kind of hardware failures, which, by the way, is rare because shaft really relies on a lot of off the shelf hardware which other companies have already productized. So our meantime, before failure, is actually quite high. It's like 180 days or so. But if there is some hardware failure, then we actually have a partnership with a brake fix company, a third party brake fix company, and these guys have 15,000 technicians all over the world. We'll air freight some parts, and often for the most common parts, we'll just have some spares on site, and then we'll fix that very quickly. So that's how the whole process works. For a partner, for robotics as a.
[00:47:49] Speaker B: Service, it's very important that to take the fear out of automation and get people used to not working with automation is really, really important as a step. So I think you're on the right path. So you've also have some news, too. You had some raises, financial raises, yeah.
[00:48:09] Speaker A: So, of course, to do AI robotics, of course, we're really leveraging venture capital, which I think is a really great ecosystem, and I think it really allows us to accelerate our growth. So, yeah, we announced around like $22.5 million in funding. And that's a combination of equity and debt. And, of course, equity is really what's helping us kind of like pay our engineers and keep the lights on and stuff. And then what we do with the equipment financing is really help finance the robots for robotics of service. You know, of course, we're paying upfront for all the hardware, but only charging our customers on a yearly basis. So, yeah, that capital, honestly, like, the vast, vast majority of that capital, like 98% or something is something like that is kind of going towards R and D, right? Like, we have a very, very small go to market team. Small but mighty. And then, you know, really, all of it's really trying to go towards making the product better, right? Because I think we believe that, look, if we can have the best product, that is kind of what, everything comes from the product, right? If it's a really good product, our current customers are going to buy more, they're going to tell their friends, they're going to tell the industry, and that's how we're going to scale, really, word of mouth, which is having the best product to market.
[00:49:15] Speaker B: Rashad, is there like an average number of robots per deployment that you have? I know you don't have a whole lot of customers, but is it like, do they all need twelve robots, or is it 20 or is it five?
[00:49:25] Speaker A: Yeah, it's a good question. So usually what we do is we start with a starter pack of six and all of our order from ones or six robots to start at least. And I think the crux of that is what we have found is with less than that, often, even if there is an roi, it's like such a small number of robots, it's hard to see Roi.
Whereas six, you can really see it, you can see the data and you can see, this is actually helping me in this and this in this way. So six is what we start with and what we do is we build in success criteria into the master service agreement, which is like, hey, look, we're going to play six to start. And this is a timeline. And of course, look, you customer have hundreds of stations that we can help with across multiple plans. You're not doing this for six robots. Honestly, that's probably not worth your time. You're doing this because you really want to leverage chef to scale. So let's work together and predefined what success looks like. So success might be like, look, I want these robots to be utilized this much time, or I want this throughput, or maybe it's like, I really want this one particularly hard ingredient, which my people really don't like. Like, for example, we've had customers who are like, risotto is just an example, but it's really very viscous. So it's really painful to do that ingredient because you're just very viscous to do with your hands. So anyways, we predefined the success criteria.
And honestly, it's kind of nice because now we know what we got to do. We can go back to our applications team, our engineering team, and be like, look, this is the requirements. This is where we have to meet. And so then we can get to work and really make sure we meet those requirements, not only just for the first six, but I also for expansion. And, I mean, the other thing I'll say with this is, like, again, like, going back to the idea of customer success. Like, it is a small industry, so we really, really try to, like, really try to make our customers happy because of this expansion thing. Right? Like, we're not trying to do this for six robots. We're trying to do this to really scale with them. So, of course, you're not going to get more robots if your first set are not doing so well. So, you know, what that means is, like, we try to have less customers, actually, but we really try to make them, like, extremely happy. Like, we'll just live there, like, just. Just to, like, whatever it takes for the customer success side to just really make them happy and really solve their pain points.
[00:51:39] Speaker B: And so what are some of the things that you need to be more successful now? Because you're going to start scaling, right? Like, now the cat's out of the bag, and we're talking about the robots and such. So what are things that you need? Are you need more programmers, vision people, AI people, marketing?
[00:51:55] Speaker A: Yeah. So I think there's a few things we need. So one is that I think this AI boom with vlms and transformers and diffusion policies, honestly, is extremely exciting. And I think what's really nice about it is the time it takes to onboard an ingredient is just getting compressed. Right. And it really allows us to accelerate that flywheel. So, yes, I think. I think we're really going to have to scale up, and we are actively trying to scale up our AI team. So I think that's kind of one really big thing we want to do.
Another really big thing we want to do is the number of deployments we can do is limited by the number of applications engineers we have. Of course, we need to grow the applications and deployments team.
What our engineering team can do is the engineering team can really help us expand to new boundaries and new customers. For example, when we started chef, we really focused on what we consider scoopable materials. If you think about a meal, 80% of it is scoopable. It's a starch, it's a diced protein. It's vegetables, sauces. Most of the meal is scoopable. But now, of course, there's oftentimes a big piece of meat, or maybe it's bread, right? Like, you want to put, like, a sandwich. You want to put bread on top, like, so we're getting into, like, piece picking of these raw materials as well. So I think that's really where our engineering team comes along. And then, like I said, our go to market team is relatively small but mighty. So we're really going to try to scale up sales and marketing as well. So I think, really, it's just. It's more people. I think, like, what's very clear now, like, abundantly clear, is we have product market fit. And now it's really, like, how do we kind of, like, leverage that product market fit to, like, continue scaling within our current partners and, of course, trying to land new partners and really scaling with them.
[00:53:34] Speaker B: No, that's great. So did we forget to talk about anything today?
[00:53:38] Speaker A: I think this is a really good conversation. I think, like, one thing that I would say that's, like, really interesting is that, you know, many of our partners. I think, you know, I think many of our partners are at least having discussions about how this labor shortage really affects them from a more strategic perspective. What I mean by this is that if you just can't hire people, it's kind of an existential crisis.
The core of production is humans today. For many of these companies, if you can't hire humans, you cannot do production. So I think for many of these companies, what we have found is, like, Roi is something they care about, for sure, but sometimes it's more just, like, strategic. We need to do this, or if we will not do this, we might not be in business in ten years. Right. Other of these companies that we talked to, they're thinking of potentially even offshoring, at least parts of their supply chain. Obviously, when it comes to food, you cannot offshore everything. There's parts that just necessarily need to be onshore, but there are certain parts you can offshore, which, of course, is not great for really any nation. But I think Covid especially taught us that relying on other nations is not great. So I think one thing that's really exciting for me, I would say, is I love robotics, I love AI, but I think there's some more macro issues that really make this very compelling, and that's really exciting for me and my team. I think that's great.
[00:55:00] Speaker B: Listen, this has been a great conversation. Great and a lot of fun.
I just had a final kind of question for you. When you're nothing automating and innovating and writing and building food robots. What do you like to do? Do you have any hobbies?
[00:55:13] Speaker A: I have a pretty simple paradigm, I think, to think about. I honestly, it sounds so simplistic, but I realize in my life I like really two things, meaningful work. I do really like building things and teams and meaningful kind of relationships. So it sounds so simplistic. But yes, I like to do certain things. I like to play tennis and go for runs and read.
But I think most of my free time that's not working is really just like with different people. I really like spending time with. And certainly we can be doing lots of different things. We'd be hiking, we could be skiing, lots of different things. But I think that's almost secondary for me. So it's really just spending time with the right people and having really good one to one conversations and relationships. I would just say simply for me.
[00:55:55] Speaker B: And how does, if somebody's listening and going, oh my God, I need to get in touch with Jeff Robotics. I need to get, I need to talk to Rajat.
What's the best way?
[00:56:04] Speaker A: Yeah, I'm Rajat at Chef Robotics AI. Or you can of course, go to our website, chefrobotics AI.
That's probably the best way. And of course, if you LinkedIn, Twitter, all the other normal resources also work.
[00:56:16] Speaker B: Thank you very much, Rajat. It's a pleasure talking to you today.
[00:56:19] Speaker A: Yeah, thank you, Jim. This is a great conversation. Appreciate you having me on.
[00:56:22] Speaker B: Our sponsor for this episode is Earhart Automation Systems. Earhart builds and commissions turnkey solutions for their worldwide clients. With over 80 years of precision manufacturing, they understand the complex world of robotics, automated manufacturing and project management, delivering world class custom automation on time and on budget. Contact one of their sales engineers to see what Earhart can build for you and their email address is infoheartautomation.com. and Earhart is spelled e h r h a righe. Our co sponsor is Anchor Danley. They are the leading manufacturer and distributor of robot bases and automation bases, high quality die sets, components blanchard ground steel plates and metal fabrication used in the production of tools, dies and molds for metalworking and plastics, ingestion and molding, mining and construction equipment and general fabrications. I'd like to acknowledge a three the association for advancing automation. They are the leading automation trade association for robotics, vision and imaging, motion control and motors, and the industrial artificial intelligence technologies. Visit automate.org to learn more and you can. If you want to get in touch with us at the robot Industry podcast, you can find me Jim Beretta on LinkedIn. Today's podcast was produced by customer Attraction Industrial Marketing, and I'd like to thank my team, Chris Gray for the music, Jeff Bremner for audio production, my business partner Janet, and our sponsors, Earhart Automation Systems, and Anchor Danley.
