“While many applications within the food manufacturing space have been successfully automated, some gaps exist when it comes to high-mix automation, including the handling of prepared sandwiches, salads, and other meals. By augmenting robots with advanced AI, automation can be scaled to all parts of the food supply chain, starting with manufacturing.”

In the automation of food manufacturing, certain challenges are inherent to the industry, particularly when it comes to handling a highly variable mix of organic and delicate items. Rajat Bhageria, president of the AI-enabled robotics company Chef Robotics, sought to tackle this challenge head-on when he founded the company in 2019. In this episode of the Manufacturing Matters podcast, Bhageria joins TECH B2B Marketing’s Jimmy Carroll to discuss Chef’s solutions for the food industry and how they can be used for difficult applications, whether it’s handling leafy greens and chicken breasts for fresh prepared meals or dealing with sticky foods like mac and cheese or peanut butter. Additional topics include the role of AI in paving the way for new automated tasks, conquering food industry automation challenges, and real-word examples of how Chef Robotics is automating food manufacturing.

Jimmy Carroll: [00:00:06] Hello everybody. My name is Jimmy Carroll. I’m the vice president of operations at Tech B2B Marketing. And welcome to this episode of the Manufacturing Matters podcast, where we discuss the trends in technology shaping manufacturing today. On this episode I have the pleasure of being joined by Rajat Bhageria, who is the president of Chef Robotics. Rajat, thanks so much for joining us. I really appreciate it. I have been following your company for a while, and I’m really glad to have you on here.

 

Rajat Bhageria: [00:00:30] Yes same here, Jimmy. Thanks for having me on. Excited for this conversation.

 

Jimmy Carroll: [00:00:33] Yeah, of course, it’s our pleasure. So I guess for those who don’t know, tell us about Chef Robotics and your company and your mission, and before you jump in, I’ll say there’s so many companies today where the name of the company might not necessarily reflect what they do. And it’s a little bit unclear, but I love how yours isn’t.

 

Rajat Bhageria: [00:00:52] It’s very clear. Yeah, so we at Chef make AI-enabled robots for the food industry, as you can probably tell. Specifically, we’re starting with food manufacturing. So that’s a little bit different than a lot of other food robotics companies. I’m sure you’ve heard of many of these that are in, let’s say, restaurant settings or even making their own restaurants that are tech-enabled. We really thought that the best go-to-market for robotics has to be manufacturing. Of course, as you very well know, robots have been around manufacturing for a while, and what we’re basically doing is adding a lot more intelligence to these robots. You know, one way we talk about this is there’s a lot of automation that works well in low-mix manufacturing, like if I’m Nestlé or if I’m Coca-Cola or PepsiCo or something like that. You know, very high volumes, relatively low mixes. And so traditional automations work quite well. Where we’re seeing a gap in the market is really high-mix automation. If you have, for example, a prepared meal manufacturer or a salad manufacturer, yogurt parfait, sandwiches. These kinds of products, these kinds of meals, basically, they have a lot of SKUs. They might have 200 SKUs. And if you have that many SKUs, you’re not going to have 200 dedicated pieces of equipment or lines. And so what you end up doing is using people.

 

Rajat Bhageria: [00:02:04] But of course, there’s a big labor shortage. And so it’s really hard to fill these roles. And it’s actually one of the biggest labor shortages in the world. And so that’s where we can come in and say, okay, well let’s make this AI-enabled robot that can be in a similar flexibility to a person. It’s obviously not the same as a person, but a similar flexibility to a person while providing the performance of a machine. So that’s  broadly what we’re doing today. The vision of the company really is to say, okay, well how do we one day get to robots in every commercial kitchen in the country in the world? And again, we really think that starting at manufacturing is the right go-to-market. Because by deploying robots in manufacturing, we’re really getting in-production training data, where our models are learning how to, let’s say, pick up cherry tomatoes and not crush them or how to pick up diced chicken without damaging it or pick up some cheese and spread it over a tray so it doesn’t clump up. We’re learning how to manipulate food basically, and if we can learn how to manipulate food really well, why can’t we use that same technology and apply it to Chipotle or Sweetgreen or Ghost Kitchen for delivery orders? So really the vision is scaling to different parts of the food supply chain, but starting with manufacturing.

 

Jimmy Carroll: [00:03:11] Yeah. So there’s so many follow-up questions there that I’d like to ask. But it’s interesting because flexibility in automation, it’s sort of an obvious term. Everybody wants a system that can handle multiple different products or even be repurposed in the future if need be. And that’s obviously not limited to the food space. And you mentioned the labor shortage. And that continues to impact manufacturers of all types. And it probably will continue to. And you mentioned food manufacturing. That’s one of the biggest gaps in terms of the labor shortage. In the food space, automation has been there. Automation has been present in the food space for quite a long time, obviously, with things like traceability and scanning and things like that or maybe more traditional inspections like looking at dented cans or something like that. But in what you’re talking about, the things you’re talking about, big challenge there. And in the last let’s say three or four years, there’s a number of folks out there identifying the food and beverage space as being sort of ripe for growth in automation. And that is a pun that I intended to use. Can you talk about the problems that you see in the market in automating some of what you’re doing in the food space and why automation and food is so challenging?

 

Rajat Bhageria: [00:04:32] Yeah, it’s a great question. So you’re right. The food industry actually historically has a ton of automation, just given the volumes. The number of products some of these manufacturers are creating is insane. And actually, a lot of other industries, for example, even Tesla, when they were building up the Gigafactories, a lot of the inspiration was actually food. Just because it’s actually relatively automated. One nuance to that I would say is that a lot of the automation is secondary packaging. So let’s say the food is in a tray. Now each of the trays or a cup or whatever, a can or what have you. Each can or tray is the same. It doesn’t matter whether it’s a pasta meal or a rice meal or what have you, or what kind of soup is inside. It’s more homogenous if you will. So you can have case packing and palletizers and tray sealers and all the other things, because it’s a relatively homogenous input. Those products and automation systems don’t really care about what’s inside if you will. There’s also a lot of, call it automation, call it machines if you will, pre-assembly. So in the cooking process and the prepping process, if you were to go to some of our customers, you would find giant pieces of equipment that can peel, for example, pineapples or something, or you have tons of leafy greens, and you want to just wash leafy greens and remove any impurities and just clean them from any dirt.

 

Rajat Bhageria: [00:06:04] Or maybe you want to take 50 potatoes and just cut them any way you might want. There’s actually a relatively decent amount of automation in the pre-assembly process and like I said the secondary packaging process. The thing that is missing I think is primary packaging. Even within primary packaging, again, it’s diving a little bit into nuances. But even within primary packaging, low-mix primary packaging is also fairly automated, right? There’s a lot of traditional kinds of dispensing, whether it’s vertical form systems or it’s pneumatic systems or gravity-fed systems or multi-headed dispensers. There’s many  kinds of augers. There’s many different kinds of feeds. There’s lots of different kinds of these. The nuance, of course, is that each of these will work for a single SKU or maybe one or two or three SKUs. And if you cut or cook the food a little bit differently, it might not be very consistent. It will still do the ingredient, but it’s not very consistent, if that makes sense. And then of course there are certain ingredients that they just can’t do.

 

Rajat Bhageria: [00:07:07] So one of our customers, they tried some of these traditional pieces of automation for something as simple as rice, but these systems would actually break the grain of rice. And that was something that they took a lot of pride in: Hey, we really don’t want to break the grain of rice. Other customers have curries with meat suspended into them, which traditional automation, it just can’t do it. Or they might be doing something really sticky, like imagine a cheesy queso or something. So basically, why doesn’t the traditional automation work in all the cases? Well, sometimes it just can’t do it, like the examples I gave. Sometimes it can do it, but it takes too long to clean it. So imagine a sauce. There’s obviously a lot of sauce dispensers. That’s a classic example. You want to put some marinara sauce on your pasta dish. Okay, great. Get a dispenser. Well, fine. But if you’re going to be doing five changeovers per day on the line, it’s not like that dispenser is running for an eight-hour shift. It’s running for 30 minutes, hour, two hours, and it’s going to take three hours to clean the thing. So probably you just want to use people, right? So sometimes it’s like, okay, you can do the ingredient, it technically can work, but it’s just not a realistic or feasible option. And the third thing is really back to flexibility. It’s just there’s too many things you need to do throughout the day, like that Station 1 on the assembly line, you’re doing leafy greens and then you’re doing a single chicken breast. Then you’re doing three cherry tomatoes. Then you might be doing some pulled pork. You get the point. There’s so much variance that, yeah, you can maybe get a dispenser to do one or two or three of those, but it’s not going to do every one of those. And so they end up using people, so high mix, broadly speaking, when you have, for example, prepared meals or salads and stuff, that’s really where we see literally hundreds, sometimes thousands of people under the same roof manually assembling meals in a 34-degree room. You know, really hard job. And that’s really where we see a great opportunity. You know, I gave all these qualifiers: It’s not in prepping and cooking. It’s not in secondary. It’s primary packaging. And within primary packaging, it’s high mix. It feels like there’s all these qualifiers. The thing with food is it’s so huge. It’s just one of the biggest industries on Planet Earth, literally, that anything you pick is a giant market. So that’s what got us excited.

 

Jimmy Carroll: [00:09:21] Yeah. When you run into so many problems too, with like organic variability, the whole ecosystem is vast in terms of what you might be touching or inspecting or handling or whatever. One  common thread that I’ve noticed in the industrial automation space, again, not just limited to food but very much in food as well, is a lot of times today when people say our system is flexible, to some extent or at least in a lot of cases that means that they’re leveraging AI. And in your case, where does AI come into play for you? Because in my opinion, it’s the idea of so-called blind robots are very useful if they’re doing preprogrammed tasks. If you add cameras, oftentimes in food it’s 3D, you’re adding 3D cameras, there’s a whole other layer of flexibility. You add AI on top of that, now you can handle all these different types of parts and products and food products too. So for you guys, where does the AI come into play and how do you think that might open new doors in food automation?

 

Rajat Bhageria: [00:10:29] I think that’s exactly right. The key innovation with Chef is truly AI. To be honest, a lot of the hardware we’re using is off the shelf, right? You know, robot arms, like you said, RGBD cameras — D as in “depth” cameras — and the CPUs, GPUs, weigh scales. So a lot of the hardware is off the shelf. We are designing a lot of our own end-of-ARM tooling, but the reason I  called this out is because you were correct that the key innovation is AI. So now why do we need that? I think there’s at least a few different layers. The crux of it is of course, like you said, the ingredients. First of all, as you correctly said, it’s organic. That in and of itself — the same exact ingredient, literally batch by batch – in the same day you’re not doing one batch. Lots of different batches. Like imagine you’re doing meat, right? Like every chicken is a little bit different. So there’s just natural variation in different parts but also different birds. So natural variation every single day it’s cut by a human. The same human’s going to cut it differently. By the way, you and I are going to cut it differently. And there’s often dozens of people cutting material. So cutting, cooking. Every single day it’s being cooked by a human, at least semi-cooked.

 

Rajat Bhageria: [00:11:42] And like I said, there’s a lot of automation, but still there’s a human partially involved. There’s gonna be variation. I mean, just to give you some simple examples. Rice is not rice is not rice per se every day. It’s a little bit more dense. It’s a little bit more sticky. It’s a little bit less and more wet. What have you. And then, of course, if you’re trying to get the right target weight, you know, speaking the same volume is not going to get you the same outcome or even sometimes using the same utensil, because if it’s extra sticky, well, you got to do something different. So that’s another example. So with all of this, AI can really help you for example in the ingredient variations. If you can have generated topography, a 3D topography of, let’s say, the pan of food, then you can get a sense of, okay, where are we? What does the topography look like? Then once you actually pick, you get some data from the robot. You get pick weight data, and you can say, okay, well let me change the policy of how the robot’s picking the next time a little bit. So that’s part one, broadly. There’s a lot more nuance in the picking. I mean, we literally have dozens, hundreds of parameters of how to manipulate every single ingredient. That is the hardest thing.

 

Rajat Bhageria: [00:12:47] Each ingredient is so variable, and within an ingredient it’s variable. But then there’s also another few layers. So portion sizes. It’s actually not as simple as you might think. It’s not like going from 50 grams of cucumbers is the same as 150 grams. There’s actually some variations, because if you have 150 grams, for example, you might need a different tool. And that tool interaction with the food is slightly different, and it might have different standard deviations, for example. So how do you still get consistency as the portion sizes vary is another hard problem. A third problem that I could talk about is, okay, now you’ve picked the ingredient, you’ve gotten the right portion size. Okay. Now how do you actually generate the trajectory such that you don’t spill the food or fling the food? Okay. So now you have to have the right trajectory generation algorithms to be able to do that. So that’s another part where AI can be really helpful. Another part is, okay, now you have to place, and when you place, every single customer has a slightly different conveyor. Some of them are multiple lanes. Some of them are not. Some of them are one lane. Some of them are sloped like literally food factories have water. They’re sloped. The floors are sloped for water drainage. So you have to place lower or higher based on the slope of the floor.

 

Rajat Bhageria: [00:13:56] Sometimes these conveyors get pushed, and so now they’re skewed. So it’s not like flush with the robot. So now you have to see that and you have to detect that. Now that could be an error, but our robots are intelligent enough to deal with that. Sometimes the operator speeds up the conveyor. And so now the velocity is just ramping or decelerating, or maybe they just stop it, hard stop. And you don’t want to just dumbly place food onto the conveyor. So there’s a lot of work to work with different kinds of conveyors and colors and slopes and heights, and without making custom software, hardware, it’s all vision-based, right? So that’s another example. The final example I can give is different trays. So even if you have different conveyors, you might have different trays. So how do you build a system that can work with clear trays and black trays and insert trays and compartment trays and non-compartment trays but you’re still having to place in different components, parts of it. So there’s just so many layers to this. It’s actually a lot harder than meets the eye. I mean even for us, we were like: Oh yeah, this seems like a perfect problem for robots. But then you actually get into the weeds and there’s so much variation in the physical world, especially with food, because it’s so inordinately variable.

 

Jimmy Carroll: [00:15:04] Oh, yeah. I mean, even the idea of bin picking, the idea of bin picking, that technology has been around a long time, but it’s still very challenging for a lot of different people not in food. And food opens up this whole new world of issues.

 

Rajat Bhageria: [00:15:24] Yes. At least in bin picking, like your iPhone is your iPhone is your iPhone. The box is the box is the box. Here, literally every day, the ingredients a little bit different. And you can be building custom stuff for that.

 

Jimmy Carroll: [00:15:35] For sure, yeah. And what I was going to say is, because I’m interested in your opinion in general industry trends too, AI is has helped some of these companies that work in warehouses, for example, be able to handle instead of just a couple of SKUs maybe hundreds or thousands or hundreds of thousands. What are some other ways, because we talk so much about AI on this podcast and I’m certain that you do too, both within your professional life and with your family and friends. Like it’s such a hot topic. Everybody wants to know about AI and what it means, and from the industrial sense in general manufacturing and logistics, warehousing, whatever, what are some other ways that you see AI really adding value these days?

 

Rajat Bhageria: [00:16:22] To be honest, I think it’s going to be used in every part of the supply chain. I think Andrew Ng, who’s Stanford professor, he’s compared AI to electricity. And I think that’s a really  fair comparison. Electricity, if you zoom out, it allowed, for example, the light bulb to happen, allowed for electric power to happen. I mean, it literally transformed every industry. It wasn’t this little silo thing. It’s like a tool that affects everything. And I think that’s 100% true for manufacturing, industrials as well. So some examples I can give. In our customers as well as most manufacturing, there’s some people at the end of the line doing QA. That’s a very manual job, and of course sometimes they can catch a defect and sometimes they can’t. And as you’ve called out, the food industry actually already has a lot of defect detection and QA systems. But they’re getting a lot more intelligent let’s just say. And being able to deal with lots of different SKUs. So now there’s a lot of these companies that are installing camera systems that have depth as well. And you can train them with AI models. And now you can do a lot more robust QA inspection, which of course increases yield and protects the inventory, but also it reduces labor or at least reallocating labor, which I think is really, really nice.

 

Rajat Bhageria: [00:17:38] I also think generally, just like data, one thing that’s   interesting. So Chef is based in San Francisco, and in Silicon Valley, data is everywhere, like every startup, every company is immersed in data. But then I visit a lot of our customers, and they’re still running on spreadsheets and sometimes even pieces of paper, literally, to plan out the production. So I just think that this data part of it is really powerful. I’ll give you a practical example from us, and I think it’s applicable. A lot of our customers, you know, they’ll have a check weigher at the end. So a check weigher might say, okay, how much weight is in the tray? Did you get the right weight? Because of course the end customer, the consumer who’s eating the meal, doesn’t want not enough rice or protein or what have you. Okay. Great. But now you’re getting the weight at the end of the tray. You don’t know if the rice is more or rather the rice was too low and the meat was too high, which of course is eating into your margins. You just have no idea. With Chef and robots, we are getting data on every single pick on every single ingredient. So it’s literally like 8–9x. If there’s eight to nine different ingredients, you’re getting 8–9x data on your lines.

 

Rajat Bhageria: [00:18:50] And by the way, that’s weight data. But you can parse it a different way, and from that you can drive uptime and plan downtime and unplanned downtime and all of these different metrics, OEE and everything else like that. So I think, okay, so that data allows you to have a lot of insights. That’s cool but it’s not crazy new as of now. Manufacturing has had data too. I think what is also cool, though, is you can now leverage that data to do more interesting things around like production planning and scheduling. Let’s optimize a production schedule for the day. I need to get this much output. I know from historic data this how much I produce, and this is the kind of yields and consistencies and outputs I can get. Let me optimize my schedule for the day, production planning and scheduling. So it’s like: Okay, Line 1 is going to run these five SKUs for this much time. I need this much input to get there. This is the allocation of human resources and people and robot resources or machine resources have been used. That’s a giant optimization problem that neural nets are really good for, for example. Humans can do it too, and traditional software can do it too, but traditional software is more rule based. ML and AI software can be a little more intelligent in coming up with that. I think there’s so many examples. We can talk about this all day long.

 

Rajat Bhageria: [00:19:58] I think, again, electricity is the right analogy. It’s going to change the fabric of how manufacturing works, because you can just have a lot more insights and data about day-to-day stuff. And then, of course, robotics is a big part of it, right? Like a lot of traditional automation has been able to solve the low-mix task. But even in car manufacturing, final assembly has a lot of humans, frankly. And I think that’s because there’s super dexterous tasks like putting the wire harness and aligning, like that’s a super, super hard task. That’s still not solved by the way. I’m not saying it’s done, but I do think innovations, for example, with imitation learning, where you can demonstrate a task instead of writing code about: Here’s how the robot should move. Imagine that the human demonstrates a task. It does this over and over. And the car harness example, the human is able to demonstrate, okay, this how you put the harness in, and now the robot can mimic it. That idea using imitation learning, which is powered by transformers, which is the same technology that’s powering Dall-E and Sora, can allow for this. So I think robotics is definitely going to be a big part of it. But generally machine learning, the data side of it, is also I think a big part of manufacturing and how AI will affect it.

 

Jimmy Carroll: [00:21:08] Yeah, that’s an awesome response. And I’m so glad that you brought up the Andrew Ng quote, because I tend to quote it on this podcast myself often, and I try not to repeat the same things too much. But I just think it’s a really good jumping-off point for a lot of people because I think that, well, let me ask you your opinion. When you’re talking to friends or family or whomever, people that are sort of outside of the general space, and you tell them that you work for an AI company, is there a fear or maybe some level of trepidation? And how do you explain that? And I know that part of it is sensationalized due to shows like “Black Mirror” or movies, “The Terminator,” and these are all awesome, but what I try to tell people is, like, “It’s a movie.” 

 

Rajat Bhageria: [00:21:58] Yes. It’s fantasy.

 

Jimmy Carroll: [00:21:59] Yeah. And AI is really valuable. So from your perspective, how do you have that conversation with people?

 

Rajat Bhageria: [00:22:06] Yeah, absolutely. And I think you’re right. I think people do have that fear. So I think what I would just say is that the world has gone through a bunch of industrial revolutions. Forget AI for a second. AI is just automation, right? So automation is not a new idea. Even the printing press is automation. Before, you have to inscribe Bibles and books and stuff, and then the printing press allows you to super quickly copy books. Well, what happened? Well, what happened is that there’s a giant influx of people who had access to books. And generally what happens with these things is when there’s lower cost, output tends to increase. If output increases, two effects happen. I think one is in the micro, and one’s more macro. On the micro level, which is to say the business level, the individual business level, that business is doing well. So they’re like: Hey, let’s print even more books. Great. So now we need to open up a second book printing factory and create more volume. So it actually creates more jobs. Our customers have the same thing, by the way. If we can actually help them increase yield, let’s say — forget saving costs or anything — just increase yield and output, they’re just going to open up more manufacturing plants. They’re going to open up more lines. They’re going to create jobs because of it. That’s on the micro sense. On the macro sense, let’s say everyone in aggregate is doing this. We’re doing this not just for Chef robots, but we’re doing this in medicine. We’re doing this across the industry, across the world. What happens is costs reduce. People are spending more. People are spending more money, right? Goods are cheaper, right? There’s more abundance, right? You can get food cheaper. You can get construction cheaper if you want to buy a house, whatever. Things get cheaper. As things get cheaper, the economy swells. You get more output. You get more GDP. As GDP increases, people feel wealthier. And because they feel wealthier, their sense of well-being actually increases. We’ve seen this a lot of times throughout history: the first industrial revolution, the tech industrial revolution. I don’t think this is anything different. These are hypothetical constructs. Bringing it home to reality, imagine social networking. Social networking was not a thing in the early 2000s, right? Facebook happened and then Twitter and then now TikTok and everything else.

 

Rajat Bhageria: [00:24:26] Now there’s people whose full-time job it is to be content creators and social media marketers and SEO optimizers. And imagine how many thousands of jobs have been created, probably millions of jobs have been created because of the advent of social networking. That’s not automation per se, but I think it illustrates the point. New technology tends to create jobs. It makes things better for everybody. Final comment I’ll say on this, at least in the case of Chef and I think many of our peers, whether it’s in construction robotics or agriculture robotics or what have you. Warehouse robotics. There’s a giant labor shortage, so there’s nobody being replaced. It’s just there’s not enough jobs. So I think all of these are fair points. And I think it depends on where the other person’s coming from. But honestly I’m not worried about this. Being in the industry, my full sense, like without any doubt, is really that it’s going to be a giant adder to the economy. I think it’s going to accelerate the economy on the micro and the macro sense.

 

Jimmy Carroll: [00:25:28] Absolutely. I think Nvidia is a good example of a company that is sort of propelling. I know that to some extent GPUs were initially launched for gaming, and they’re in computers, general purpose computers. But these days GPUs are almost synonymous, at least in the industrial space, with AI, machine learning. Nvidia in Q4 of last year had a record quarterly revenue of almost $40 billion, which is up 78% from the year before. So if you want a talking point of the growth of AI, there it is.

 

Rajat Bhageria: [00:26:06] Exactly, exactly.

 

Jimmy Carroll: [00:26:08] So we’ve talked a lot about some applications of where Chef is being used. And I know that largely in this space we’re bound by NDAs. But are there any examples of customers that allow you to talk openly? I’d love to hear some. I bet you have some fun ones.

 

Rajat Bhageria: [00:26:25] Yeah, for sure. So I can give you a few examples of what  kind of benefits our customers have seen. So one of the ones that I really like a lot is one of our customers called Chef Bombay. They’re based in Edmonton, Canada. And this is the customer actually that I alluded to earlier. They had tried a lot of traditional automation for dispensing rice. It didn’t work because they cut their rice. It broke the grain. And that wasn’t okay for them from a quality perspective. And so we decided to partner with them. And the president was thrilled because of a few different ROIs. One is that we made it really clear: There’s a labor shortage. We’re not firing anybody. But now you can reallocate people from the assembly line to do other, better tasks, better as in perhaps even more humane. Doing manual work all day, doing this motion all day for eight hours a day is honestly physically extremely taxing, especially because it’s a cold room. So now those people are freed up. Two is we were able to increase the average throughput for their lines by around 20%. That is huge because now they have 20% extra capacity.

 

Rajat Bhageria: [00:27:31] So they work with some fairly big retailers as customers. They’re a contract manufacturer. They can literally just produce more volume. So whatever the cost of Chef might be, they more than made that up in spades because of the increased revenue and volume. And I think, by the way, just zooming out, I think some of the best businesses are the businesses that create revenue more than they cut costs. So like Uber, for example — again, not a manufacturer but just to illustrate the point — Uber creates revenue for people. You or I can drive Uber, and we get money. And I think the best businesses create revenue more than they cut cost. I think that’s a good proxy, so for this customer, we were able to create revenue for them, which is awesome. And then finally we were able to increase yield for them, actually save them on food wastage and things like that, and that’s meaningful in any manufacturing space, but especially food, because 50% of revenue goes to raw materials. So I think that’s a good example of some of the ROI that we’ve been able to generate.

 

Jimmy Carroll: [00:28:29] This something I like to ask about. Seven or eight years ago, nine years ago, there was this common thread in the industry or maybe in the world that robots are here to take jobs. And it was all these reports that come out basically saying, well, yes, but no. And in fact, not really. Like a lot of times it’s creating jobs. And now it’s almost like people have gotten over that fear of robots. But now there’s this fear of AI. Well, you’re at a company that’s at the cross section of robotics and AI. And you’re saying the same thing: No, we’re not necessarily taking jobs. We’re giving people jobs that are no longer — I know it’s cliche — dull, dirty, dangerous and letting them go do something more valuable. And I’ll bet that your technology, at least high level, is designed so that it’s relatively easy to use. So that somebody who used to work on that floor can now operate an automated machine. Is that right?

 

Rajat Bhageria: [00:29:33] 100%, 100%. And honestly, that’s something we think a lot about. I think there’s a lot of automation that’s honestly — I’m sure you’ve used it — it’s really quite hard to use. You need to be technical or relatively technical and trained. We take a lot of inspiration from consumer apps that you and I use every day, like Netflix. Everyone can use Netflix. We take a lot of inspiration from very simple user experiences like TikTok and Netflix and Facebook even. So yes, exactly. We designed our user experience to be extremely simple. It’s like, okay, I want to run production. What mill am I running? What ingredient am I running? Okay, it’s pre-populated with the portion size. The robot’s going to say: Hey, you need to get this utensil. Get the Cs50 utensil. It says it right on the HMI, on the dashboard. Plug in the Cs50 utensil, put in your pounds of food, press play, run. That’s it. You teach somebody once for five minutes, and then they’re off to the races. And we think a lot about how do we actually, for example, do things that make this even simpler. So, for example, the entire dashboard is in Spanish, which is a language that many people in the food industry use. We can obviously do any language you might want. It’s mostly photos. Even the text is limited.

 

Rajat Bhageria: [00:30:44] It’s mostly photos or videos or GIFs and things like that. Exactly like you said, those line workers are now becoming robot operators. They’re becoming continuous improvement managers. These people know the business. Some of these people, like at Amy’s Kitchen, one of our customers, they’ve been with the business for 20, 30 years. I mean, they’ve been with Amy’s let’s say for 20 or 30 years. So they deeply know the business, but it’s just that this job is redundant. Previously Amy’s had no choice. It’s like: We don’t have a choice. There’s no automation that can do this. We just need people to do this. But now I think people are like, okay, well, if Chef robots can do the manual assembly work, now we can repurpose those people and actually let’s optimize things. Let’s increase throughput. Let’s figure out how to do better material handling and material flow. Let’s figure out how to do better QA. Let’s build other systems and processes and reallocate that labor. Let’s open up a new line. That’s the kind of thing we’re seeing. And these people are now upgrading themselves, up-leveling themselves to where they are now robot operators, which is awesome.

 

Jimmy Carroll: [00:31:49] Yeah. And there’s this funny irony there too. Automation can help businesses protect against the labor shortage, but in some ways, and I’m not going to make the comment that automation can do this fully, but in some ways it can help bring in some of the human talent, because if I were a young person and I had no job, and I’m looking for a job opening and there’s two jobs. One of them is packing eggs. And the other is operating a robot that packs eggs. I’m going for that robot job.

 

Rajat Bhageria: [00:32:25] One hundred percent. Absolutely. And I think one of the things we try to do is really make the line associates feel really bought in. This not a threat. It’s a tool. It’s literally a tool. We’re going to teach you how to use the tool. So we’ll do a lot of small things that I think are quite meaningful. So whenever we ship a robot, we do a little naming ceremony. Hey, let’s all circle around the robot. Not a Chef team. The line associates. What name do you want to give the robot? Of course we do a full training, multiple trainings, and then, yes, from a hiring perspective, it’s been a great thing because now it’s like, one of the first line items is “Robot Operator” or at least “Interact with robots.” Even if you’re not operating the robot, you’re around the robot, you’re learning about it. You’re learning what works and what doesn’t work. And that’s very useful.

 

Jimmy Carroll: [00:33:23] Yeah. Yeah. This really cool. We’ve talked about a lot, and there’s a million questions that I could ask, but like I said, I want to be respectful of your time. Is there anything we haven’t talked about that you feel would be important to mention today just in general?

 

Rajat Bhageria: [00:33:38] I think as a general trend, zooming out. This is again a very macro statement, To your point about young people, I think a lot of young people, my sense is they don’t want to do manufacturing. They’re more attracted to, after high school, after college, let’s do everything else. Let’s get a job at Uber. Let’s try Uber. Let’s do DoorDash, what have you. And I think that’s of course not great for the U.S. I think the U.S. really does need — manufacturing is a core strength. Historically what’s happened, as you know, is we’ve off-shored quite a bit, and I think what COVID really taught us, frankly, like many other countries, is how fragile our supply chain can be if we don’t have production on-shore. And that’s not just for food. I mean steel and semiconductors and iPhones and all the rest of it. So especially as we are seeing that labor shortage in general manufacturing, I think robotics and AI is critical because we do need to keep production on-shore, which is critical to national security.

 

Rajat Bhageria: [00:34:49] And so I would just encourage young people: Manufacturing is really cool. And I think especially as you add AI and robots, it’s actually really powerful. And I think there’s a lot of hype in AI robots right now, about robots in the home and robots in restaurants and robots in hospitals. My sense is that’s more hype than reality right now. Even Chef — we are in food robotics. We could have gone to restaurants, but we decided to go to manufacturing. By the way, all the food robotics companies that went to restaurants are now dead. It didn’t work. And we can talk about why that is, but robots aren’t totally ready yet for the day-to-day world, and I hope they get there not too far in the future because of all these advancements in AI. This moment robots can be very successful in manufacturing. So if you’re young and you want to do AI robots, manufacturing is the place to be. So I would just encourage people to think about that.

 

Jimmy Carroll: [00:35:45] Yeah, absolutely. The other thing to think about is even if you’re not going in as an engineer, if you go into welding, for example, there’s some stat out there about all the skilled welders in the U.S. are going to be retiring in whatever amount of years it is. I don’t remember the stat off the top of my head, and automation has taken over. But there are these jobs too, where not just the people that are building robots or developing AI, but you can go and make a really, really good career, whatever it might be, welding, working on the plant floor or otherwise. And I just think, yeah, it’s important to reframe the idea of working in manufacturing as some sort of fallback, because it’s absolutely not.

 

Rajat Bhageria: [00:36:30] Yep. Exactly.

 

Jimmy Carroll: [00:36:32] Rajat, thank you so much for the time. It’s been super interesting to talk to you. I really appreciate the time. If people want to learn more, I think it’s chefrobotics.ai. Is that right?

 

Rajat Bhageria: [00:36:42] Yep. That’s correct.

 

Jimmy Carroll: [00:36:44] Yep. And everyone can check them out on LinkedIn and follow the latest happenings in events and cool case studies and videos and everything. Again, thanks so much. I really appreciate the time. If anybody has questions for Rajat, I’d be happy to take those. You can reach out to us via manufacturing-matters.com. And thanks for listening. And Rajat, once more, thank you so much.

 

Rajat Bhageria: [00:37:04] Thank you so much, Jimmy. Super fun having this conversation.

 

Jimmy Carroll: [00:37:07] Of course. My pleasure.