Chef Robotics Hits 10M Meal Milestone in Under Two Years. The Secret? AI-Powered Robots Trained With Lots of Field Data

This week, food automation startup Chef Robotics told The Spoon it had reached the ten million-meal milestone, less than two years after the company’s first robot was deployed in June 2022.

If you think that type of growth is achieved by steady month-over-month increases in production over time, you’re wrong. In fact, according to founder Rajat Bhageria, after taking nearly a year to reach its first million, the company’s been on an up-and-to-the-right full-throttle ride of hockey stick exponential growth ever since. The next million took about one hundred days, the next after that three weeks, and, nowadays, Bhageria says it takes just about two and half weeks or so per additional million food items.

That’s a whole lot of meals made in a short time, which made me wonder what type of customers and food facilities the company serves with its robotics. Bhageria says their typical customers are those running centralized food processing facilities, where Chef Robotics assembles the type of yogurt parfait or protein platter SKUs you might pick up at your local coffee shop. Other end-products include airline meals, hospital food service, and other pre-packed meals. In other words, Chef Robotics’ robots aren’t making salad bowls or pasta from a menu in a restaurant, but instead assembling pre-packed meals at high volume across a wide variety of food types.

“If Tesla’s core technology is batteries, our core technology is food manipulation,” Bhageria said. “Which is to say, we need to be able to go from shredded chicken to diced chicken, to cubed chicken, and from julienne onions to chopped onions to sticky cheese grits in marinara sauce. The whole point of Chef Robotics is to be as flexible as possible.”

Bhageria says their food assembly and manipulation systems differ from traditional dispensing systems, which are limited by a hardware-centric approach and lack sensitivity to the variability in food ingredients. He says Chef Robotics focuses much more on software while leveraging a combination of computer vision, motion planning, and a robotic arm equipped with various utensils. This technology integration mimics the dexterity and intelligence of human food handling, enabling the robots to adapt to different ingredients and recipes rapidly.

Chef Robotics’ CEO says that the company has essentially overcome the infamous AI “cold start” problem by accumulating a massive amount of data by having its robotics in the field enable it to become more flexible over time and understand the different challenges around different types of food manipulation. He says this has resulted in what they call its (what else?) ChefOS.

“ChatGPT can basically just download the Internet,” Bhageria said. “But there’s no training data for food manipulation. You can’t really do it in the lab because the way one customer, julienne’s an onion, is very different than the way another customer, julienne’s an onion. The way to learn how to manipulate food is you have to deploy robots.”

The more robots you have in the field, the more data you have, which makes the AI smarter, which means the existing robots work faster and faster; in other words, it’s the virtuous circle of scaled automation and AI.

When I asked Bhageria the names of his specific customers, he told me they aren’t disclosing who they are because the companies using Chef Robotics technology wish to remain under the radar for now. We do know that the company has robotics in five cities across the US and Canada and plans to triple its fleet of robots this year.

With this type of growth, it won’t be long before Chef Robotics’ robots are pumping out a million meals assembled in just a matter of days.

Leave a Reply

Your email address will not be published. Required fields are marked *