Somewhere in a lumber yard in the Pacific Northwest, a robotic arm reaches for a piece of plywood. It tilts. It hesitates. It senses a knot, a warp, the slight give of a wet board. It adjusts grip pressure by a few newtons and sets the board onto a moving conveyor without splintering an edge. The whole motion takes under two seconds. Nobody on the floor looks up. That, more or less, is what RIOS Intelligent Machines built.
RIOS spent eight years convincing factories to give their robots a sense of touch. Not a metaphor. An actual electrical sense - tactile sensors, machine vision, edge AI, and a containerized brain that could decide, in real time, what to do with a board, a bottle, a tray of frozen dumplings. The company was founded in Menlo Park in 2018 by five engineers out of Xerox PARC, the research lab famous for inventing about half of what your office computer does. They left to start something more physical.
The bet was simple. Industrial robots had spent forty years being strong, fast, and blind. They moved heavy things from point A to point B with magnificent indifference. But anything unstructured - a piece of wood, a piece of fruit, a soft package, a stack of mixed cartons - broke them. RIOS thought the answer was not a bigger arm. It was a smarter hand.
Bernard Casse, the CEO, and Clinton Smith, the CTO, assembled a team that wrote CVPR papers in the morning and shipped FANUC integrations in the afternoon. That is rare. Most robotics startups pick one. RIOS tried to do both, and for a while it worked beautifully.
By the Numbers
The 2024 Series B was a $13 million round, co-led by Yamaha Motor Corporation and IAG Capital Partners. Yamaha had been an investor since 2020 and doubled its stake. Main Sequence doubled its pro-rata. That kind of follow-on math is the closest a venture round comes to an endorsement signed in blood. The plan was to push deeper into three verticals: wood products, beverage distribution, and packaged food.
None of those are sexy. That is the point. Founders who chase glamour build a humanoid that pours coffee at a trade show. Founders who chase revenue build a workcell that pulls misaligned plywood off a saw line at 3 a.m. RIOS, to its credit, was in the second camp.
The product line read like a small operating system for the physical world. AI Perception & Computer Vision handled real-time inference, segmentation, and 6DoF pose estimation. Robot Control & Motion talked to FANUC arms, tracked conveyors, and spoke PLC. Edge AI Platform ran containerized agents on the factory floor with a no-code logic builder for plant engineers who did not want to write Python. Mission Control, launched alongside the Series B, was the cloud layer that let one operator watch dozens of workcells across multiple sites. Underneath all of it: tactile sensors that RIOS engineers had been refining for years.
The customers were not famous. They were the kind of companies that make the materials inside the things you buy. A lumber mill in one state. A beverage warehouse in another. A frozen-food plant somewhere in the Midwest. The robots showed up, learned the job, and stayed. By the time RIOS published its final website update in early 2026, the company counted more than 25 edge sites and 8 FANUC robots running in customer production.
And then, quietly, it stopped. The rios.ai homepage now reads "2018 - 2026" and offers the company's intellectual property for licensing and open source release on GitHub. There is no press release. There is no public obituary. There is just an archive page and a list of capability areas, written in the careful past tense of an engineer cleaning up the lab.