Skip to Content
MIT News feature

Engineering lunch

MIT’s “Spyce Boys” bring robotics to the kitchen to deliver fast food that’s both healthy and cheap.

At 10:29 a.m. on May 3, Kale Rogers ’16 gave his hands a final rub with sanitizer and unlocked the door of Spyce for its grand opening. Spyce, a fast-casual restaurant in downtown Boston founded by Rogers and three other recent MIT grads, bills itself as the world’s first restaurant featuring a robotic kitchen that cooks complex meals to order—and it has multiple patents pending to support that claim.

The gaggle of people waiting outside filed in, whipped out their cell phones, and began taking photos and videos of the open kitchen and the touch screens for placing orders. They posed for selfies with COO Rogers and his cofounders, CEO Michael Farid ’14, SM ’16, lead mechanical engineer Luke Schlueter ’16, and lead electrical engineer Brady Knight ’16, who all smiled somewhat nervously as they welcomed their first customers.

As customers placed their orders, Spyce’s automated food storage bins (known as hoppers) reliably fed refrigerated ingredients through a portioning system that measures precise quantities into a red box that zips along a horizontal track. That box, called the runner, collects ingredients and delivers them to one of seven induction-heated woks that spin to tumble the food so it cooks evenly at 450 °F. The hoppers look like simple funnels, but Rogers says there’s “some stuff in there” being patented, and developing them “took a lot of iterations.” Completing the visible technology are screens, situated above the woks, that give customers progress reports on their meals.

The founders’ inspiration for this business came naturally. All four studied mechanical engineering at the Institute and were members of the water polo or swimming teams. Being athletes, they had hearty appetites—but, Rogers says, “we wanted fast, cheap, and healthy food, and we couldn’t get all three together.”

So in 2015, when Rogers and Knight were juniors and Farid was working on his master’s, they signed up for StartMIT, an intensive entrepreneurship workshop offered during Independent Activities Period (IAP). Their goal, they agreed, would be to prepare fast, delicious, healthy food that someone on a student’s budget could afford. (Schlueter joined the team that spring.) Each of the four founders had previously taken the robotics and product design classes 2.007 and 2.009; three of them had taken 2.12 together, co-­developing a soccer-ball-kicking robot. Maybe it’s no surprise, then, that they soon built their first cooking robot, working in the basement of what is now the Theta Tau fraternity house.

A photo of Spyce's four founders
Spyce cofounders Brady Knight ’16, Kale Rogers ’16, Michael Farid ’14, SM ’16, and Luke Schlueter ’16

The original IAP prototype, Knight says dryly, “only worked in the videos.” The hopper mechanism, in particular, needed tweaking. The IAP version, which used a pizza slicer attached to a motor to release ingredients into a wok, “just sprayed quinoa everywhere,” he says.

When the team built a second prototype that summer as part of the Martin Trust Center for Entrepreneurship’s summer-­long accelerator program (then known as the Global Founders’ Skills Accelerator and now called MIT delta v), they drew on their experiences in class. “In 2.007,” Knight says, “we got our first grasp on how to use rapid prototyping techniques to iterate on a robot design.” That would prove critical because they needed to develop their first truly functional prototype by the end of the summer for the accelerator’s Demo Days showcase—and before classes and water polo season started in the fall. Among other things, they solved the spraying-quinoa problem by ditching the pizza slicer and coming up with a new way to hold and portion ingredients. But the new prototype was too big to fit through the door of the basement in which it was built. So to get it over to the accelerator showcase, they had to disassemble it, haul it to Lobby 13, and put it back together. When they reassembled it, everything worked, says Knight: “It was a miracle.”

Even so, their cooking robot was still a work in progress. That fall, it dumped half-cooked food straight onto the counter in front of a potential investor. One outcome of that fiasco: each of the restaurant’s automated woks now has a sensor telling it whether there’s a bowl underneath. Eager to keep improving their prototype, Knight and Rogers took 2.671 as seniors, which taught them how to conduct thorough experiments. In that class, Knight tested ways of cooking chicken that helped the team fine-tune their drum-sautéing technique, and Rogers studied induction from an electrical standpoint to get a better handle on the induction coils they’d built into their design over the summer. In the spring of 2016, Spyce won a Lemelson-MIT Student Prize (in the “Eat It” category) and the MIT $100K Entrepreneurship Competition’s audience choice award.

Even as they were perfecting their automated kitchen technology, the founders knew they needed more than technical expertise to develop a successful robotic restaurant. So Farid got in touch with restaurateur Daniel Boulud, the chef-owner of multiple award-winning restaurants and author of nine cookbooks, by guessing his e-mail address in five tries—and the team ultimately convinced him to serve as Spyce’s culinary director and invest in the concept. Boulud then introduced them to Sam Benson, who had worked for him at the Michelin-starred Café Boulud in New York and had done research and development at Chipotle. The “Spyce Boys,” as the four jokingly call themselves, were thrilled when he signed on as executive chef.

Once Benson came onboard, Rogers explains, they could work on their technology and the menu in tandem, developing functionality that would allow the robotic woks to execute the chef’s culinary vision. For instance, to meet Spyce’s nutrition goals, Benson wanted to make a lot of use of cooked kale. Kale is chockablock with vitamins, calcium, antioxidants, and other nutrients—but it’s fibrous and tends to clump together, and to a robot it’s fragile, requiring more careful handling than other ingredients fed through the hoppers, like rice and chicken. As lead mechanical engineer, Schlueter spent three months solving the kale problem, tinkering with the hopper yet again to reduce the risk of jamming and fluff up the chopped kale so it would sear, not steam.

“We originally had a different hopper design for kale,” Knight says. But they found that the hopper they created to deal with the peculiarities of kale worked better all around: “We ended up changing the design of all the other hoppers to be almost identical to the kale hopper design in the end.”

To point out another advantage of developing the menu and the robot together, he invokes a past example of food-prep automation that failed. A quarter of a century ago, Taco Bell introduced its Automatic Taco Machine at a test “restaurant of the future” in Irvine, California. The thing could crank out 900 tacos an hour. But it broke down a lot, and a company spokesman said customers felt the machine-made tacos lacked the “human touch” they were used to. So the test was discontinued and the disgraced machine was shipped back to headquarters, where it has sat, unused, ever since. While Spyce’s kitchen is robotic, its recipes reflect the very human and creative touch of an innovative executive chef.

A photo of a finished Spyce dish
Spyce’s Thai bowl.

Sustainability and clean tech, which uses resources as efficiently as possible, are also important to the founders. So chicken and salmon are on the menu, but not beef, because it’s significantly less sustainable. The serving bowls, cups, cutlery, and napkins are all compostable or recyclable, and the kitchen is self-cleaning in order to minimize the amount of water needed to operate the restaurant. A sharp spray of hot water cleans each wok after each use. On average, the robotic kitchen uses about 0.3 gallons per minute, or roughly 20 percent as much as the average commercial dishwasher.

Spyce has a test kitchen at Greentown Labs, an incubator in Somerville, Massachusetts, for startups with a clean-tech focus. And it has commissary and food-prep space in Malden, Massachusetts, where humans wash and chop the fresh ingredients overnight. The three sites together employ dozens of people. “Automating away workers is not our angle,” Knight says. Rather, he contends, Spyce has offloaded a lot of drudge work from humans to machines for the sake of efficiency, and he says the remaining human jobs pay better than typical fast-food jobs.

Indeed, at the restaurant, staffers are stationed between the row of woks and the serving counter, adding cold ingredients and garnishes to the bowls of robot-cooked food. And a human greeter is on hand to help customers with the touch-screen ordering kiosks, which allow them to customize any of the seven basic bowls—Latin, Thai, Indian, Moroccan, Chicken + Rice, Hearth, and Lebanese—by adding and subtracting ingredients such as avocado, freekeh, and microgreens.

Customers often spend more time dithering at the touch screen over which bowl to order and what to add to or subtract from it than Spyce takes to cook the meal. Within three minutes of inserting your credit card in a kiosk, you’re handed your bowl. A standard bowl—which includes three or four optional garnishes—will cost you $7.50. The robotic kitchen can make up to 200 meals an hour, about as many as more conventional fast-casual restaurants produce.

Photo of customers using Spyce's touchscreen menu-exploring and food-ordering computers
Photo of Spyce employee dressing a bowl of food with guacamole

Less than three minutes after customers order, an employee adds garnishes to their robot-cooked meal.

And how’s the food? Over the course of three visits, this reporter couldn’t find a single customer willing to say anything bad about it. A five-year-old girl and her mother, a pair of elderly men, and women getting takeout to eat at work all gave it enthusiastic reviews. Even a 17-year-old who had previously rolled her eyes at the menu (“It all looks so healthy!” she’d said, not meaning it as a compliment) dug into her Chicken + Rice minus the pomegranate seeds—and ate every bite, proclaiming, “This is actually really good.”

By a month after its opening, the restaurant had served more than 10,000 meals. Business remained brisk. Benson, the chef, was working on “menu R&D” (developing seasonal menus, for example). The Spyce Boys were building a mobile app for advance order placement. And Jimmy Fallon had made a joke about the restaurant on Late Night. “It was cool that we were on his writers’ radar,” Knight says.

When asked what made these four mechanical engineering majors think they could develop a commercially viable robotic kitchen, he has to stop and think. “Ten years ago, you couldn’t have done this—the tools we used are fairly new,” Knight says, referring to the actuators and controllers that allowed them to quickly prototype the Spyce kitchen and build it cheaply enough. “When we started this, three and a half years ago, there wasn’t any buzz in the food industry about automating restaurants,” he says. Instead of trying to ride or even create a trend, they just focused on figuring out how to make affordable and nutritious fast food. “Being engineers, we came at the problem from a new angle,” he says. Then he chuckles and adds, “If I had known how much work it would be, I don’t know if I would have tried.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.