The Robot Will See You at $14 an Hour
Tutor Intelligence just raised $75M to rent out robots by the hour, with a human on call for when they get confused. Finally, a workplace where the new hire is actually expected to ask for help.
A basement, a welder, and the occasional elevator outage
Most robotics origin stories begin in a garage. Josh Gruenstein's began in the basement of his childhood apartment building, where he welded robots and, more than once, took the building's elevators down with him. The neighbors, one assumes, were early and involuntary investors.
His co-founder, Alon Kosowsky-Sachs, became known at MIT as the guy with the homemade robotic Segway, the kind of campus reputation that works as either a warning or a recruiting pitch. At MIT the two started Tutor Intelligence, which builds general-purpose robot hardware and the AI that runs it, with the declared aim of taking robots where they haven't gone before.
This week Tutor announced a $75M Series B led by a16z, with USV and Neo joining. The company now has 160 people and robots at work in 20 states, in Fortune 50 factories and in small family businesses alike. Its fleet has quadrupled in a year, is growing about 20% a month, and has helped make and ship more than 250 million products. If you bought groceries this week, there's a fair chance a robot touched something in your bag before you did. It did not, we're told, read your receipt.
The robot will see you now, for $14 an hour
Gruenstein's sales pitch has the brisk economy of a classified ad: "Hey, have you ever considered a robot for 14 bucks an hour?"
The premise is that robots needn't wait for perfect autonomy before earning a living. Tutor rents them out by the hour. When one gets stuck, say on a box at an awkward angle or a pallet with opinions, a remote human, the "tutor" of the company's name, takes over and guides it through. That rescue becomes training data, and the whole fleet gets slightly smarter. It is an apprenticeship in which every mistake is instantly shared with several hundred siblings, which is either the future of learning or a description of group chat.
The reason this matters is unglamorous and important. Language models learned from the internet; robots have no internet. No one has posted a few hundred thousand clips of how their grip shifts as they lift a crate of canned tuna. Tutor's bet is that deployment itself produces that missing archive, while customers get real work done in the meantime. Somewhere, a chatbot trained on forum arguments is feeling a little underqualified.
For buyers, the arithmetic is tidy. Robot hours are billed like shifts, so they land on the books as operating expenses, the way human labor does, and deals typically close 40 to 60 days after a first call. Gruenstein describes the result as a win three ways: cheaper goods (he likes to say Tutor shaves about 10% off everyone's Cheerios), line workers moved off the most punishing tasks into better jobs, and an industry that finally gets the real-world data it has been starving for.
Cassie, Sonny, and a room full of robots practicing
Tutor owns every layer of the business: it trains its own robot models, designs and builds its own machines in its own factory, and deploys them directly to customers. Rather than customizing a robot for each client, it ships general-purpose units that reach a new site within weeks and are usually working within a day of arrival, occasionally within a few hours.
Two robots do most of the lifting, in both senses:
- Cassie moves pallets and other heavy loads, running on models of 200 to 400 million parameters.
- Sonny, an industrial semi-humanoid with dexterous arms, picks and packs on a 4.5-billion-parameter model.
Gruenstein reckons the pair can handle 80 to 90 percent of the work in manufacturing and logistics.
Back at headquarters, 100 robots spend their days in a room called Data Factory 1, which Tutor believes is the largest robot data factory in the United States. It can produce up to 10,000 hours of training data a week. Code ships to the fleet daily, with major releases and manufacturing changes every two weeks. The robots, in other words, receive more frequent performance reviews than most of us, and complain about them less.
Factories are only the opening act. The company talks about construction, about recovering materials from landfills, perhaps one day about homes. Founding team member Emily Spector's view, roughly: the technology works now and the problems exist now, so waiting for some grand AI breakthrough is a strange way to spend the time.
The science fair is every third Friday
Tutor is based in Watertown, Massachusetts, about 20 minutes outside Boston. The office includes what the team calls a robot playpen: a mock e-commerce warehouse where new models are tested before any customer meets them.
The rituals are charmingly nerdy. An internal lunch-and-learn series lets anyone present on something unrelated to their job; past topics have included computational painting, prime numbers, and the case that 18th-century New England whaling was the ancestor of venture capital. One imagines the room, at a VC-backed company, was receptive. Every third Friday there's an internal science fair, where engineers and researchers print posters about their last three weeks of work and everyone wanders through. This being a robotics company, some of the posters move.
The Head of People holds a PhD in organizational psychodynamics, a detail that suggests the company takes the human side of building machines seriously.
Who thrives there? Tutor says it wants people who love the game, including the hard parts and the roadblocks: the sort who meet something absurd and get excited rather than tired. One person often owns what would be a whole team's job elsewhere, and some of the best work comes from tiny crews of an engineer, a researcher, a salesperson, and someone who knows the customer cold. The job also gets you out of the building. You might spend the morning on an equation at a whiteboard and the afternoon at a tuna factory, watching your code pick things up. Gruenstein calls it walking and chewing gum; the robots, presumably, are still working on the gum.
Who they're hiring, and why you might raise your hand
Tutor is hiring across the stack: software, hardware, electronics, materials, and technicians who assemble the robots. Several priority roles are company "firsts," the rarest kind of opening, where you define the job rather than inherit it. And plenty of them aren't engineering at all. Robots, it turns out, also need marketing, finance, operations, and someone to recruit the humans.
Start with the firsts. Tutor wants its first Head of Marketing, someone to own category, brand, and demand generation across both mid-market and enterprise sales, reporting straight to the CEO. It wants its first Controller, to keep a manufacturer's balance sheet honest: inventory, bills of materials, landed cost, warranty. And it wants a Head of Company Operations to set the operating rhythm for every function, also reporting to the CEO and sitting on the leadership team.
Then there are the people who keep the machine, and the machines, running. A Technical Customer Engineer would look after a portfolio of live customer sites, diagnosing whatever goes sideways, whether software, mechanical, or process. A Technical Recruiter would hire the engineers and researchers as the team grows past 160. A Senior Technical Program Manager would run the Cassie program and, fittingly, automate away the manual status-tracking that program managers everywhere quietly dread.
On the technical side, much of the work can be done from Boston or San Francisco, which is good news for our Bay Area readers. A Senior/Staff Software Engineer in Robotics would lead motion planning, real-time control, and perception, while a Senior/Staff Software Engineer in Platform would shape the core systems, from backend services to ML infrastructure, that make the rest of the company faster. A Research Scientist and a Senior Research Engineer would build the AI behind new robot capabilities; Tutor says it will shape the research role around the person, and weights the engineering one toward strong programming fundamentals over any particular stack. Finally, in Boston, a Head of Hardware Engineering would lead mechanical, electrical, firmware, and manufacturing engineering and set the hardware roadmap with the CTO.
The full list lives on Tutor's jobs page.
A note for our community: robotics is still a field where many of us end up being the only woman on the factory floor, or at the whiteboard. That is precisely why rooms like this one need more of us. If you've ever wanted a job that is half equation and half tuna factory, this may be it. And if the robots give you any trouble, remember: they are, by design, still learning.