How AI Is Changing the Future of Food (Chef Robotics Explained)

Manav: Hello everyone, today we have Rajat Bhageria on my show. He's the founder and CEO of Chef Robotics. Rajat, I'm so excited to have you on the show — how are you doing today?

Rajat: I'm good, and thank you, Manav, for having me. I'm excited.

Manav: Thank you. I want to ask you many questions about Chef Robotics, but I really want to ask about you first — can you give us a little intro about yourself, starting Third Eye, moving on to being an investor, and then starting Chef Robotics? How has the last decade of your life been?

Rajat: That's a big question. I guess a lot of my schooling was in Cincinnati, Ohio. In high school, I was very focused on one goal — I wanted to get into a good school, so I worked really hard toward that. Once I got accepted, I felt like, okay, I've spent the last five years laser-focused on this one thing, now I'm free, I can do whatever I want — and that's really when entrepreneurship started for me.

Senior year of high school, I worked on a few different projects. One was called Café Mocha — at the time, Medium didn't really exist yet, there was Tumblr, some WordPress blogs, but I thought, I like to write, and there wasn't really a good platform for young writers to publish their writing, research, or poetry to the world. That's really where I learned software engineering, actually shipping something. I was 16, 17 at the time, growing up in suburban Ohio, and suddenly there were people in Brazil, Argentina, China using this product — it felt pretty incredible, and it was a formative experience for me as a young person. But that was really my first real entrepreneurial endeavor.

Then, freshman year of college, my goal was actually to find a co-founder for Café Mocha. I was doing computer science at the time, thinking, okay, who's the best engineer around here? I found some genuinely great people who became close friends. What happened next was we entered a hackathon, PennApps. This was right around when AlexNet had happened, and computer vision felt like it was having a real resurgence with deep learning, CNNs. We thought, why not take that and combine it with smart glasses — Google Glass was having its moment too — and build something for the visually impaired, just for the hackathon.

Manav: How exactly were you helping the visually impaired with the glasses?

Rajat: It's funny — there were three of us, and one of my co-founders at the time had a grandfather who was visually impaired, so that's really where we understood the problem. The idea was that visually impaired people often develop this sense of "learned helplessness" — going through life constantly needing help for small things. Our thinking was: what if we gave them a product they could wear on their face, smart glasses, and they could give a verbal command, like "okay, glass, recognize this" — and we'd capture a picture or video stream and run real-time object recognition and detection. Then we could tell them verbally, "you're looking at ibuprofen," or "that's a dollar bill" or "a five-dollar bill," or read them the restaurant menu.

Manav: That's such a good idea.

Rajat: It seems simple in hindsight — honestly, we didn't know much about computer vision at the time, we were like two weeks into college. But we ended up doing really well at that hackathon, just hacked our way through it, and at the end, a few angel investors said, "why don't you turn this into a company?" We thought, why not, what do we have to lose? That became Third Eye, and a huge chunk of my college experience ended up being that company — we worked on it for around three and a half years. Eventually it came down to: drop out and go full-time, or sell the company. For various reasons, we decided to sell.

Manav: You must have learned a lot about image recognition and machine learning through that process too.

Rajat: Exactly — that was really my first hands-on experience with AI, and I got genuinely excited about the whole space. This was around 2018, 2019. Personally, I thought, okay, this seems exciting, and I had two things on my mind for what came next. First, I really wanted to learn how to build companies from someone more senior, an actual mentor. Second, I wanted to figure out what to build next. I tried to run both in parallel.

On the learning side, I convinced Slava Rubin, founder of Indiegogo, who was giving a talk at Wharton — I introduced myself, told him about Third Eye. He was launching a new equity crowdfunding product at Indiegogo at the time, and I pitched him: "you probably need help with deal flow, what if I help you build a new pipeline using scouts around the country, basically other founders?" I was essentially trying to create my own job with him, and in exchange, I asked if I could just follow him around, attend the meetings he attended. He said yes. For a few months I just followed Slava around — went to all his meetings, he was thinking about acquiring companies, I helped build models, made slides for board presentations, pretty intense stuff for a young person to have access to, honestly, but he became a genuinely great mentor.

That was the daytime; in the evenings, I was thinking a lot about what to do next, and got excited about two ideas — AI, obviously, and energy. It felt like those two things, in different ways and scopes, were going to be the two biggest forces shaping our lives. Given the Third Eye experience, I leaned more toward AI, and started really digging into: who are the actual customers here, what's the right market?

I got excited about the idea of "AI in the physical world." At the time, most AI companies were doing things like spam detection, or Netflix's recommendation engine — mostly cloud-based, purely software. I thought, there's something exciting about AI applied to the physical world, since well over 90% of GDP happens in the physical world.

Manav: Like Tesla, Figure, humanoid robots.

Rajat: Exactly, that kind of thing. My thinking was: what are the foundational industries that make up most of GDP? It's the labor market — that's roughly half of GDP. Retail jobs, nursing, hospitals —

Manav: Transportation, construction, even mining.

Rajat: Right, these industries are just enormous. So I thought, if I want to work on the biggest slice of the physical world, that's the labor market, which naturally points toward AI-enabled robots. I still didn't know exactly which industry, though — and that's actually when the impetus for the fund came about. A close friend of mine, Nandeet, had done something similar to what I did with Slava, except with a bunch of LA-based funds — basically saying, "I'll do whatever you need," which meant helping close deals, but also fund operations, whatever it took.

Manav: I can tell you really enjoyed that experience — the investing, looking at deals.

Rajat: I did. We became close friends and decided to launch something together — Prototype Capital. The idea was: there's going to be a big wave of companies applying AI, IoT, ML to the physical world, and we wanted to go after them — but these companies don't just live in the Bay Area, LA, or New York, they're everywhere, Cincinnati, Idaho, Atlanta, all over the US and the world. Traditional Silicon Valley VCs weren't really looking there. So how do we get access? By finding out who founders actually hang out with for fun — and the truth is, most founders hang out with other founders. If I want access to a great insurtech company in Atlanta, hypothetically, I can find that founder's friends, convince them to become a scout for us, offer carried interest on deals we invest in through their referrals, and hopefully surface some genuinely great, under-the-radar deal flow.

That's basically what we did — we ended up with around 70 different founders scattered across the country acting as scouts, and we invested in a bunch of companies through that network. From Chef's perspective, it also put me in front of a lot of potential customers, in construction, agriculture, food, all sorts of industries — and I ended up focusing on food.

I always knew Prototype's core idea, founders investing in other founders, was going to be a thing I kept doing, but personally, I really love building — I never wanted to become a full-time VC, at least not in the short or medium term.

Manav: You're too young for that anyway.

Rajat: [laughs] Right, I really wanted to build. So the plan was, Prototype stays a side thing, and the full-time focus becomes building. The food industry ended up being genuinely exciting to me, for a couple of macro reasons. One was just the sheer size of the industry — once I'd narrowed toward "something in food," I spent a lot of time looking at data on industry size, and learned that the single biggest labor industry in the US, by headcount, is nursing and personal aides. Second is retail sales. Third is food preparation, food service, food production.

My read was that the first two aren't really tractable for AI anytime soon, which meant, of the industries actually addressable by AI, food might genuinely be the biggest one — using headcount doing that job as a rough proxy for AI market size. So I thought, okay, this is a genuinely big industry — but is there a real problem to solve? Because obviously, in startups, you need an actual problem.

I took both an anecdotal and a data-driven approach. Anecdotally, I talked to food truck operators, fast-casual operators, airline catering companies, ghost kitchens, manufacturers, everyone — and they all said essentially the same thing: "my number one problem is a huge labor shortage. On any given day, I don't know what percentage of my staff will actually show up, and because of that, I'm leaving revenue on the table." That felt like a real pain point anecdotally, and I confirmed it with data — the Bureau of Labor Statistics reported in 2023 that the food industry has the number one labor shortage in the US, worse than retail, worse than manufacturing, across assembly, food prep, everything — basically the entire food industry.

So those were the two things that got me excited: it's a big market, and there's a real, confirmed pain point. And one more thing — it felt technically plausible that AI robots could handle scooping food into a Chipotle-style or Sweetgreen-style bowl. We hadn't done a ton of deep technical homework at that point, but the pieces felt like they fit together, and that's why we ended up focused on this broad space.

Manav: Can you talk a bit about what these robots are actually doing? In my opinion, the hardest part is that food is genuinely hard to manipulate, because it's so variable.

Rajat: Today our go-to-market is really food manufacturing, which surprises a lot of people — most assume "robots for restaurants." But as you noted, we're actually focused on manufacturing. Think about any meal you might get on an airplane, or a frozen meal from the grocery store, or the prepared salads in the deli section at somewhere like Trader Joe's — all of that is made by people, in large facilities, essentially food factories.

The way it typically works: you have long assembly lines, maybe 12 people on a line, each with a big tub, scooping food into trays, burritos, wraps, sandwiches — the most mundane, repetitive task imaginable. Humans shouldn't be doing that long-term — there's no future where humans keep doing this. So that's what we focus on: food assembly. That means scooping food out of a big hotel tub without crushing it, hitting any portion size the customer specifies — say, exactly 53 grams of shredded chicken, consistently — detecting, tracking, and placing ingredients like carrots into the right compartment, spread exactly the way the customer wants, and doing all of this in a way that's genuinely scalable, not custom software or hardware built per ingredient, per tray, or per customer, but a flexible, AI-driven solution.

Manav: Let's talk business. An average assembly-line worker might make $50 to $60K a year, roughly. How do you price your service — is it a yearly subscription?

Rajat: We charge a yearly recurring fee, which comes in below the cost of the human labor it replaces — each robot is roughly doing the work of two people. And to be clear, nobody's being fired — those people move to a different task, usually something less repetitive and more valuable. Our model includes a small implementation fee up front, what we call an NRE, a non-recurring expense, mainly covering initial configuration and installation — we send applications engineers out to deploy the system and train the team. There's no big capital expenditure required up front. Once that's done, they pay us a yearly recurring fee, which is lower than the cost of the two people it effectively replaces.

But honestly, cost savings is probably number five on the list of reasons customers actually care. The bigger ROI drivers: there's a genuine labor shortage, so a lot of facilities can't even run all their production lines — say they have 10 lines but can only staff seven, so they're under-producing relative to customer demand. If we can say, "put eight robots on that idle line and now line eight can run," that's real new revenue for them. We think a lot about revenue increase specifically. We can also usually raise average throughput — robots don't get tired, don't slow down six hours into a shift the way people naturally do — so Chef often increases average throughput, which again means more revenue. Generally, I think the best businesses increase revenue more than they save cost. We also help with yield — less food gets wasted.

Manav: What's next for Chef Robotics — are you raising another round, what will you be focused on over the next year or two?

Rajat: A few things are top of mind. First, we have a genuinely strong set of existing customers, quite large ones, with plants all over the world — so a lot of our focus is landing and expanding within those accounts, which is really about customer success, essentially living alongside them, making the product extremely good for their specific operation. There's a lot of product and engineering work behind that, but we invest heavily in it, because if a customer buys two robots, that's not that impressive — but if that same customer buys 50, that's genuinely impressive, because they wouldn't scale up like that unless it actually works. That recurring revenue per robot per year is the sweet spot, good for them and good for us.

Second, we've only recently really come out publicly with what we do — we were very quiet for a long time, deliberately, because we felt like we were onto something, and we were so focused on serving current customers well that adding more sales or marketing headcount wouldn't have done much — frankly, we couldn't even handle more demand at the time. Now we can, the product feels ready to scale, and we have the team to execute, so a big focus now is scaling go-to-market, sales, and marketing to bring on new customers.

Third, we're continuing to invest heavily in AI. We now have a dedicated AI team using imitation learning, learning from demonstration, and decision transformers — using production data combined with imitation learning to learn new SKUs, new products, new ingredients, working toward a more generalized food manipulation model.

Manav: From researching you and talking to you, you're really good at picking the right problem — I think a lot of people, myself included, have been kind of unintentional in choosing what ventures to pursue. What advice would you give people on deciding what to focus their time on?

Rajat: That's obviously a big topic. Broadly, I've heard two schools of thought on this. One is the "passion hypothesis" — you look inward and ask, what am I passionate about, and you go do that. That can definitely work, especially if your passion happens to line up with a big market and a real opportunity.

But honestly, before Chef, before Prototype, I wasn't inherently passionate about visually impaired accessibility, or about food specifically — I love robots and AI, but I never particularly loved cooking. I wasn't deeply passionate about the specific domain going in, and I think a lot of my founder friends are the same way — Aaron Levie wasn't waking up passionate about cloud storage, or HR software, most people aren't inherently passionate about these specific domains.

There's an alternative hypothesis, where the causality runs the other way — success leads to passion, not the reverse. I tried the passion approach first, asking what I genuinely cared about, and landed on "I like AI and robots" — but then I flipped it, and asked instead: what's the right company to start that has the highest probability of actually succeeding? Because once something starts going well, it becomes very easy to become passionate about it — if you're genuinely good at something, if you're winning, the passion follows naturally.

For me that meant: find a genuinely big market — look at the data, what's the biggest addressable market, which pointed to labor, and within labor, food was the biggest tractable slice. Then, is there a real problem — yes. Then, what's the right go-to-market within food — assembly, not cooking or prep. Then, manufacturing specifically. If you look at Chef today and ask, "why food manufacturing," it sounds unglamorous, kind of boring on the surface — but if you trace the actual history behind it, it makes complete sense, and it's a foundation for where we build next. So I tend to think: figure out what's actually going to make you successful first, and let the passion follow from there.

Manav: Well, with that said, thank you so much for coming on the show — you answered everything amazingly.

Rajat: Thank you so much. Yeah, thank you, Manav, I appreciate it.