Metabob revolutionizes AI code analysis and optimization through innovative applications of cutting-edge technology. In this insightful session, Dave Vellante of SiliconANGLE Media hosts Axel Lönnfors, chief operating officer at Metabob, at the Rosewood for theCUBE + NYSE Wired event. Lönnfors discusses advancements in AI code analysis, providing a glimpse into Metabob's use of graph neural networks to streamline code optimization and refactor substantial legacy systems.
The Metabob platform leverages AI by integrating graph neural networks with large language models, effectively modernizing and detecting anomalies within extensive codebases. Co-hosted by theCUBE Research, the discussion explores how Metabob’s capabilities assist companies, ranging from government agencies to Fortune 500 firms, in managing their technical debt. Lönnfors details the enterprise-driven approach and the journey towards achieving product-market fit.
Key insights from the conversation include the importance of accurate anomaly detection and automated fixes for maintaining operational efficiency. Lönnfors emphasizes Metabob’s unique position, highlighting its focus on preserving code context to prevent issues such as 502 errors. They assert that customer satisfaction and value delivery remain the company's guiding principles, steering Metabob towards greater integration into AI-driven development environments.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Physical AI & Robotics Leaders. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Physical AI & Robotics Leaders
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Physical AI & Robotics Leaders. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Physical AI & Robotics Leaders
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Physical AI & Robotics Leaders. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Rajat Bhageria, Chef Robotics
Metabob revolutionizes AI code analysis and optimization through innovative applications of cutting-edge technology. In this insightful session, Dave Vellante of SiliconANGLE Media hosts Axel Lönnfors, chief operating officer at Metabob, at the Rosewood for theCUBE + NYSE Wired event. Lönnfors discusses advancements in AI code analysis, providing a glimpse into Metabob's use of graph neural networks to streamline code optimization and refactor substantial legacy systems.
The Metabob platform leverages AI by integrating graph neural networks with large language models, effectively modernizing and detecting anomalies within extensive codebases. Co-hosted by theCUBE Research, the discussion explores how Metabob’s capabilities assist companies, ranging from government agencies to Fortune 500 firms, in managing their technical debt. Lönnfors details the enterprise-driven approach and the journey towards achieving product-market fit.
Key insights from the conversation include the importance of accurate anomaly detection and automated fixes for maintaining operational efficiency. Lönnfors emphasizes Metabob’s unique position, highlighting its focus on preserving code context to prevent issues such as 502 errors. They assert that customer satisfaction and value delivery remain the company's guiding principles, steering Metabob towards greater integration into AI-driven development environments.
Rajat Bhageria, founder and Chief Executive Officer of Chef Robotics, joins theCUBE hosts at the NYSE CUBE Studios to discuss the role of AI-powered robotics in transforming the food industry. This installment is part of theCUBE's ongoing series in partnership with NYSE Wired, spotlighting trends in physical AI and robotics.
In this engaging session, Bhageria shares their expertise on how Chef Robotics addresses labor shortages in the food industry through AI-enabled robots. The discussion, led by theCUBE Research analysts, explores the company’s uniqu...Read more
exploreKeep Exploring
What is the role of AI-enabled robots in addressing labor shortages in the food industry?add
What is the strategic approach being discussed for entering the food manufacturing industry?add
>> Hello, I'm John Furrier, host of theCUBE here at our NYSE CUBE Studios, the NYSE and theCUBE, partnering with NYSE Wired program, building out a new community of trusted leaders. Part of our physics and physical AI series. Physics is the big buzzword in meals with robotics, AI technology. Rajat Bhageria is here, founder and CEO of Chef Robotics. Back on theCUBE, recently was part of our AI Infrastructure Robotics Leaders series in June, Rajat, great to see you again. Thanks for coming to our New York studio, NYSE Wired CUBE here. Thanks for being part of the team.
Rajat Bhageria
>> Thanks for letting me.
John Furrier
>> Talk about, set the table, Chef Robotics, what you guys do, Brian Baumann had a nice demo at your facility a couple months ago and the momentum you have.
Rajat Bhageria
>> Yeah. So what we do is basically make AI enabled robots for the food industry. The big pain point in the food industry, as with many industries, is there's a big labor shortage, 1.1 million people, that these companies are looking to hire, only expect to get worse. And so basically what happens is that traditional automation, which is kind of dispensers, depositors, doesn't work. So they end up having to rely on people to do this redundant work. We build robots that can help automate the assembly part of that, which is actually 60 to 70% of labor. So in terms of momentum, we've now made, I think, 73 million servings, which is awesome. That's more than essentially all the other kind of food robotics companies essentially combined, which is really cool. And one thing that's interesting about Chef is that we're starting not in restaurants, but rather in manufacturing. So today, if you go to Trader Joe's, or Whole Foods, or Costco, or what have you, and you get a frozen prepared meal or fresh prepared meal, odds are that's actually made by Chef, which is pretty cool. The vision of the company is that if we can actually use this training data about how do you manipulate rice and not clump up, how do you manipulate a grape and not crush it, how do you manipulate sauce so you spread it over a tray? If you can learn how to manipulate food in a factory, then that training data is training data that you can kind of apply all over the place.
John Furrier
>> So Grab 'N Gos, we have been there.
Rajat Bhageria
>> Yes.
John Furrier
>> Whole Foods, that's my lunch, a little sushi here, salads.
Rajat Bhageria
>> Correct.
John Furrier
>> And then also it's large volume coming in.
Rajat Bhageria
>> Yes. Super large volume.
John Furrier
>> So is there a certain kind of food that's use cases out of the gate that are more aligned with what you guys are doing?
Rajat Bhageria
>> I think one thing that we realize is most food, if you think about it, is we call scoopable, which is to say you scoop it with an ice cream scoop with your hands. So even a yogurt parfait or burrito, yes, you have to fold the burrito, but inside, it's all scoopable stuff. Most stuff on a sandwich is scoopable. So that's where we started. So we've done like 2000 plus ingredients. Most of it is in a tray, but now we've kind of more recently done wraps, burritos, sandwiches, things like this, like an egg sandwich type of thing. The next kind of thing we're kind of expanding into is more like what's called piece picked food, which is like, I want one chicken breast. I own one salmon filet, I want to make one burger patty, things like this. And that kind of is the universe of food basically. Food is either scooped or piece picked.
Dave Vellante
>> So what is the state of the art of robotics today for at least in your sort of use cases? Because are you building learning systems as opposed to this sort of standard, see, think, and act, rules-based system?
Rajat Bhageria
>> Yes.
Dave Vellante
>> Can you explain where we are today?
Rajat Bhageria
>> Yeah. It's a great question. So maybe I even take it even before all the AI stuff. The robotics as of old has been around for 40, 50 years.
Dave Vellante
>> Sure. Yeah.
Rajat Bhageria
>> I mean, if you go to a Ford factory, you see lots of robots. These robots are essentially hard coded though. They do this motion all day long and they're very good at it, but they have very few sensors, they're very few, very little software. And so those kind of pieces of equipment work very well for what's called a low mix manufacturing. What low mix essentially entails is like, there's one line and line one is going to do this product all day long. So there's the Ford F-150 line, it's going to do F-150s all day long, or there's the AirPods Pro Line, that's just going to do AirPods Pro all day long. It's not going to make an iPhone. That's kind of a solved problem. Now that's still not easy, by the way. There's a lot of work that goes into that, but it's essentially like there's systems integrators who are very good at doing this. What's changing in the world, of course, is that manufacturers and people, consumers want choice. In the food world, you don't want the same Stouffer's dinner every single night, right? You want choice. So what's happening is this explosion in SKUs products. So in the food industry, instead of having five products that Coca-Cola does, you might have 500 products. You have vegan meals and meat lovers meals and gluten-free meals. And each of those sectors has its own 50 SKUs.
So if you have 300 different SKUs, you're not going to have 300 custom dedicated lines you're going to pay half a million dollars for it. That's not a good economic idea. Instead of what you do is you get flexible lines. These lines kind of go from doing one meal to the other meal to the other meal all day long. So I think that's really where we're seeing, okay, well, we can actually leverage more AI and machine learning to take essentially all the existing hardware. Let's not try to reinvent the hardware stack. Let's take collaborative robots, let's take LGBT cameras, let's take NVIDIA GPUs, Intel CPUs, et cetera, et cetera, and let's add AI and ML to make these robots flexible enough to go from doing diced onions to peas to mashed potatoes to whatever throughout the day.
John Furrier
>> But they set up the runs though, right? It's not like it's intermixed or is it-
Rajat Bhageria
>> They set up the runs.
John Furrier
>> And so that's where the simulation comes in, that's where we're seeing the digital twins and the software side of it.
Rajat Bhageria
>> Yes.
John Furrier
>> So your multipurpose on the robotics-
Rajat Bhageria
>> Yes.
John Furrier
>> Dave, I always say this on theCUBE Pod, it's like the matrix. Upload how to fly a helicopter. In a way, the robots are programmable.
Rajat Bhageria
>> Yes.
John Furrier
>> And then the lines go.
Rajat Bhageria
>> Yes.
John Furrier
>> You service the order.
Rajat Bhageria
>> Yes.
John Furrier
>> They wait.
Rajat Bhageria
>> Yes.
John Furrier
>> New update comes in, so it's pluggable programming.
Rajat Bhageria
>> Correct. Yeah. And exactly. Our customers are getting new customers. Their customers are changing what meals they want every single week, month, honestly. So they're constantly saying, "Okay, instead of this meal, I want to do this meal." And so they can very quickly onboard new ingredients and new meals. And we have this really nice VLM based system where they can actually say, "Okay, well, I want to do this new ingredient, upload a photo and basically we'll predict the robot motion that the system should take with that new ingredient."
Dave Vellante
>> And by choice of the name you chose suggests you're going to be in this industry for a while.
Rajat Bhageria
>> Yes.
Dave Vellante
>> I wonder if you could explain that, the market dynamic, because you're not trying to build general purpose consumer robots that fold my laundry.
Rajat Bhageria
>> Yes. Yes. Yes.
Dave Vellante
>> You're really focused on this industry.
Rajat Bhageria
>> Yes.
Dave Vellante
>> What's the thinking there?
Rajat Bhageria
>> No, it's a really good question. I think robotics is really hot right now, but when we started in 2019, it was a bit of a winter and there was like a slew of dead robotics companies. You probably know many of them.
John Furrier
>> We've seen them all. A lot of a dead bodies.
Rajat Bhageria
>> A lot of dead bodies. And so my thinking was, okay, look, if I'm going to do ... I think everyone agrees on the vision of robots. We all agree with that. But the reality at the time, at least in 2019, was stark. So I was like, okay, if I'm going to do something, I need to be very practical. And so trying to build a general purpose system was technically not really tenable, frankly. And I would even argue today, this idea of a general purpose system is probably-
Dave Vellante
>> I mean, they're out there.
Rajat Bhageria
>> Out there. Yeah.
John Furrier
>> The matrix will come soon where you have more agility in the robotics.
Rajat Bhageria
>> Yes. Yes.
John Furrier
>> But you guys already have fine-tuning or you have a robotics that could pick a grape.
Rajat Bhageria
>> Yes. Yeah.
John Furrier
>> So you got advancements in the hardware, software.
Dave Vellante
>> You said you were going to dominate a niche and this is a big niche.
Rajat Bhageria
>> Exactly. So the thinking of Chef is like, okay, look, start in food manufacturing. That's an industry that's a giant industry all over the planet. Everyone in the world has prepared meals. Okay, start there. What happens there? A few things happen. The most important thing is we learn how to manipulate food. Training data is training data, right? If I can, like you said, pick up a grape, I can do that in Sweetgreen, I can do that in Travis Kalanick CloudKitchens, right? So we start there. We can actually get revenue. We get operational backbone about how do you manufacture robots. We get operational backbone on how you service robots. We reduce our bill of materials by getting volume with our suppliers. We get case studies in the food industry. And the most important thing is this training data. And then over time, we go from like high volume kitchens to lower and lower volume kitchens. So the next kind of step for us is like commissaries, kind of central production kitchens, but they're not factories. So imagine airline caterings, central commissaries, imagine CloudKitchens, that's kind of step two. Step three is kind of the day-to-day kitchen, right? Has casuals, prisons, hotels, stadiums, event venues.
John Furrier
>> Talk about the business model. And Dave and I were talking before we even started doing interviews about Michael Dell and his innovation around supply chain. He would put the suppliers near his factory so they'd be close. Obviously, food is perishable.
Rajat Bhageria
>> Yes.
John Furrier
>> You're in San Francisco. Distribution of the food is important. What's your plans on that? And two, you mentioned CloudKitchens. Are they a customer or a distributor? Or take us through the market selection of one, you're thinking around the distribution, orders, servicing them, and then partners.
Rajat Bhageria
>> Yeah. So it's actually interesting actually, even from our customer's perspective, by the way, this is not exactly what you're asking, but it's an interesting kind of point to make and I'll answer your question. A lot of our customers are actually thinking of offshoring parts of their supply chain because they can't have enough labor. And obviously the current administration policy is not helping this. So I mean, that's also kind of a strong why now, like why robots need to exist? We can't outsource food. That's not good for the US by any sorts of imagination, right? But to answer your question, honestly, most of our customers are not in San Francisco or something. We're there because of talent. There's the best AI engineers, ML engineers, robotics engineers. That's why we are there. But our suppliers are all over the country and the world. I mean, we get our robots from Denmark. We get a lot of parts from Germany. We get some parts from China, kind of globalized supply chain. And then our customers are also all over the US and Canada, usually outside of major cities. We have some customers in New Jersey. We have some customers in upstate New York, Idaho, Oregon. It's kind of like where you'd expect kind of factories.
John Furrier
>> And you're shipping the food from San Francisco?
Rajat Bhageria
>> No, no, no. We ship the robots.
John Furrier
>> Oh, you ship the robot. Okay, that's what I'm saying.
Rajat Bhageria
>> Exactly. So we build these robot modules and these robot modules are the same kind of footprint as a person. We put them in crates, ship them out to our customers.
Dave Vellante
>> CloudKitchen deals with the food, right?
John Furrier
>> So they buy your robots or at least, whatever you're making, you do it.
Rajat Bhageria
>> Yes. So today we're actually not doing CloudKitchens. Today we're doing factories. Today we'd be going to the guys-
Dave Vellante
>> That was the next step.
Rajat Bhageria
>> That's the next step. CloudKitchens is the next step. But yes.
Dave Vellante
>> But your customers are dealing with the food.
Rajat Bhageria
>> Yes, exactly.
Dave Vellante
>> You're providing the infrastructure.
Rajat Bhageria
>> And by the way, I think that's an interesting point. There's been a few different robotics companies that have tried to vertically integrate, which is to say they try to do the robots, they try to have the brand, they try to have the food. What we have found is that doing all three is quite hard. And my thinking is, look, Chipotle or Starbucks or what have you, or Whole Foods, they're very good at food. They've been doing it for decades. They're very good at it and they're very good at brand. We're good at robots. And by the way-
John Furrier
>> Stick with what you know.
Rajat Bhageria
>> Exactly. And by the way, the probability of robotics, somebody succeeding is still very hard. It's a technically extremely intense company. The probability of restaurants succeeding, I would argue is even smaller. Now you're stacking those probabilities on top of each other, which is why a lot of these companies that have kind of vertically integrated have perished. So my thinking has always been, look, robotics is very much a B2B thing, in my opinion. I think you sell to people who are very good at that end thing, and then you're very good at the AI robotics, hardware manufacturing, things like this.
John Furrier
>> Yeah. And they get benefits because they see cost reductions, top line growth, better product.
Rajat Bhageria
>> Yes.
John Furrier
>> I mean, it's just a win. All right, what's the hardest technical thing that you guys are working on now? Because obviously it's very fast-paced.
Rajat Bhageria
>> Yes.
John Furrier
>> Physical AI is the hottest thing.
Rajat Bhageria
>> Yes.
John Furrier
>> And they changed our name of the series to AI Robotics Leaders to Physical AI because what we're talking about here is physical AI, digital and physical coming together, which is the first party relationship between software and data.
Rajat Bhageria
>> Yes.
Dave Vellante
>> But robotics is cool.
Rajat Bhageria
>> Robotics is cool.
John Furrier
>> And we have robotics in the name. Physical AI and robotics.
Rajat Bhageria
>> Yes.
John Furrier
>> Because robotics arc is here. I mean, it's finally here.
Rajat Bhageria
>> It's here. It's here.
John Furrier
>> It's not winter.
Rajat Bhageria
>> It's real. It's real.
John Furrier
>> Yeah.
Rajat Bhageria
>> Right. I think it's actually a lot harder than you might imagine. So there's this law in robotics called Moravec's law. And Moravec's Law basically says that the things that are very easy for humans, like picking up this cup are exorbitantly hard for robots. And the things that are really hard for humans like linear algebra are really trivial for machines and robots. So I think just to kind of put a very precise example, doing something as simple as like, how do you pick up rice, which is a very simple ingredient, cooked rice, jasmine rice. And how do you get 45 grams or whatever gram the customer wants, 10 million times, consistently, every single time is actually really hard because every single time the topology of the pan of rice is changing. And by the way, humans are cooking the rice. So every single today, it's a little bit more wet or less dense or what have you. So we need a lot of sensors to say, okay, dynamically, here's how you pick from every single time to manipulate it. That gets harder by the way with meats. So imagine a tub of shredded chicken. Well, by the way, if you go through 10 million servings, that's a lot of different birds you're going through, which means every one of them is different. There's different density. So basically dealing with organic stuff is actually really hard.
Dave Vellante
>> And you're saying because your example of a grape before without squashing that grape because as a human, it's very easy to-
Rajat Bhageria
>> Yes. It's very trivial.
Dave Vellante
>> ... not squash the grape.
Rajat Bhageria
>> Yes. Yes.
Dave Vellante
>> Because you have that tactile feel.
Rajat Bhageria
>> Tactile. Yes.
Dave Vellante
>> Okay. So how do you solve that problem? Is that software, hardware, integration?
Rajat Bhageria
>> It's a little bit of all of it. So I think that that's like one of the key kind of insights with Chef. One thing that we really rely on is this idea of a utensil. It sounds like a very simple idea, but what the humanoid companies are trying to do is build a general purpose hand. But if you look at kitchens, it's basically all utensils. You have scoops, you have spoons, you have tongs, you have lots of different utensils. That makes the problem a lot more tenable for us. And that's not trivial, right? You have to design the right utensil and then on top of that, you still have to figure out what pose basically, where in the pan of food do you enter from. Then we have a pneumatic system to figure out like the pressure and the flow for the pneumatics that you pick from. So anyways, there's a bunch of different things that you have to kind of get right. And then it's like you try a bunch of stuff, that's where the learning comes in. You try and see what happens and you try more. You're kind of playing with the food until you kind of converge in a policy that allows you to kind of get all the different aspects you need right. So it's not just like picking food or not. It's like, first of all, can you pick? Then it's like, can you get the right portion size? I want 80 grams. Okay, great. Then can you be consistent around the 80 grams? Then do you spill the food? Then do you crush the food? Then can you do this fast? Because customers want speed. Then can you do this reliably? Then can you do this not just for that particular ingredient, but the billions of other ingredients? That's what's hard.
Dave Vellante
>> So today is the volume business. Can you take us through the business case for each of these sort of targets?
Rajat Bhageria
>> Yes.
Dave Vellante
>> What does it look like?
Rajat Bhageria
>> It's a great question. So our customers really usually approach us because they can't hire the people. So the way they're thinking about this is, okay, line five over there, I can't run line five because line five requires 20 people. I don't have the 20 people.
Dave Vellante
>> Okay. So it's not a cost replacement, that's a revenue example.
Rajat Bhageria
>> Revenue increase.
Dave Vellante
>> Yeah. Okay.
Rajat Bhageria
>> So the number one thing is revenue increases. Now to be clear, there's some factories where labor is not an issue. They're in some part of the country where labor's not an issue.
Dave Vellante
>> So are you a fit there?
Rajat Bhageria
>> Well, then there's different ROI. So in those cases, we can usually actually help them increase throughput.
Dave Vellante
>> Okay. So it's still a revenue.
Rajat Bhageria
>> And I think by the way, the best businesses, in my opinion, are the ones that help other people make more money. It's not about saving costs. It's about making money. So number two is throughput. Number three is we can usually help them with yield, which is to say food giveaway. The idea there is a simple one which is people on average tend to give away or put too much food into the tray. And that is a really meaningful amount of money for these food companies because 50% of revenue goes to raw materials. So if I can help them decrease the amount of giveaway by 2%, that is a huge amount of money for them. So that's three. And the fourth thing is labor. Labor is number four though. So it's interesting because a lot of people have this perception of like, oh, robots are taking jobs. No, there's not enough jobs. None of people want to do work in a 34 degree-
Dave Vellante
>> We need robots.
Rajat Bhageria
>> We need robots.
Dave Vellante
>> Because we don't have enough people.
Rajat Bhageria
>> There's not enough people.
Dave Vellante
>> And then those economics flow through to the other two sort of domains that you're targeting?
Rajat Bhageria
>> Yes. Yes. I believe so. I think labor shortage is going to be generally something that persists across even Sweetgreen. Why is that? Because frankly, there's much better jobs out there now. I'd rather do DoorDash or Uber or whatever else. I get to work on my own hours. I probably make more money than being in a very hot or very cold kitchen.
Dave Vellante
>> Yeah. I was listening to a pitch yesterday from the VC event I was at and the firm, their whole value proposition was basically providing staff for restaurants.
Rajat Bhageria
>> Yes.
Dave Vellante
>> And I'm like, I don't like that business. First of all, you're going to have a hard time finding people.
Rajat Bhageria
>> Yes. Yes.
Dave Vellante
>> And it's okay. So you got a limited supply of people.
Rajat Bhageria
>> Yes.
Dave Vellante
>> Uber drivers, that's not a problem but people would work-
Rajat Bhageria
>> Yes.
John Furrier
>> Autonomous vehicles are coming.
Dave Vellante
>> People are involved too. They're working... For the period of time, plenty of people wanted to drive Ubers, but you can't find people for restaurants.
Rajat Bhageria
>> Yes. Yes.
Dave Vellante
>> That's their biggest problem.
Rajat Bhageria
>> Yes, that's a big problem.
Dave Vellante
>> So robots are such an obvious solution, so I didn't like that pitch.
Rajat Bhageria
>> Yeah. Yeah.
John Furrier
>> Talk about... To close out the segment, I wish we had more time, but I definitely would love to see you tour when I'm in San Francisco.
Rajat Bhageria
>> Yeah. Of course.
John Furrier
>> What you're working on now, put a plugin for the company, share some stats. What are you guys looking to do? Obviously hiring is key.
Rajat Bhageria
>> Yeah.
John Furrier
>> You're working with the GPUs, there's some tech involved.
Rajat Bhageria
>> Yes, yes.
John Furrier
>> Share some stats, momentum. What are you looking to do? What are you optimizing for?
Rajat Bhageria
>> Yeah. So I think in terms of what we're working on, I think there's a few different things. One is usually our customers are pretty big customers, like enterprise, like Amy's Kitchen, for example, they started with a couple robots, now they're 20 robots. So really investing in success and making those customers successful so they buy more robots. That's like a big focus for us. The second thing is we are expanding obviously within continental America and UK. So kind of go to market is a big focus of mine. The third big thing that we're really focused on is saying, "Okay, well look, at this point we made 73 million meals in production. I mean, we have a ton of trading data." So how do we leverage a lot of these modern techniques like imitation learning and learning from demonstration combined with the developments with transformers and say, "Okay, let's leverage human demonstrations for novel ingredients, plus the 73 million demonstrations from production data to make a really great, kind of more generalized foundation model for food."
John Furrier
>> I like the training data because in this rapid accelerated evolution, synthetic data has been talked about because there's lack of real time data.
Rajat Bhageria
>> Yes.
John Furrier
>> You're getting real data.
Rajat Bhageria
>> Yes.
John Furrier
>> Computer vision, sensor data, mechanical data.
Rajat Bhageria
>> Yes, yes. And I think by the way, that's a really key point because even if you look at Waymo and Cruise, like Waymo, I think anytime I see Waymo, it's such an incredible experience. It's like magic. But if you look at Waymo, of course they do a lot of SIM, but they have had kind of cars in the roads of San Francisco for years now, three, four years. I mean, even deep into COVID. So they have tens of millions of miles of in production kind of training data. So I think honestly, part of the industry is over hyping SIM. I think the reality with robots is like there's so many edge cases, edge cases you can't even imagine. It's just like the unknown unknowns. And you're not going to know that unless you try stuff. I'll give you an example. In our industry, we've had cases where we're like, okay, we're placing onto a conveyor, relatively constrained environment, but there's humans around us. So we see cases where the human bumps into the conveyor and now the conveyor is skewed. And so now the robot's like, "Oh, I didn't know about that." So now we have to build software that autonomously detects that. And that's just one example. There's thousands of these examples, even in our case, which is honestly relatively constrained, which is why we started in the first place.
John Furrier
>> Yeah. Yeah. That's why I love the three scaling laws that Jensen talked about last week. There's training, reinforced learning inference and then deep thinking because now you have the data coming in.
Rajat Bhageria
>> Yes.
John Furrier
>> The data mode is phenomenal. So congratulations.
Rajat Bhageria
>> It's the biggest mode. And I think with any robotics to me, that is mode.
John Furrier
>> Rajat, thanks for coming back on. Again, we'll see you soon. You're in the fold. Thanks for being part of our community at theCUBE and NYSE Wired.
Rajat Bhageria
>> Yes. Thank you.
Dave Vellante
>> Great to have you.
Rajat Bhageria
>> Yeah. Thanks for having me.
John Furrier
>> Great to have you on and congratulations on your success.
Rajat Bhageria
>> Thank you.
John Furrier
>> Robots are cool. AI is cool. We're doing our best to be cool and bring you all the data here with theCUBE. I'm John Furrier, Dave Vellante. Thanks for watching.
John Furrier
Dave Vellante