London- and Sydney-based Reimagine Robotics emerged from stealth on August 3, unveiling a platform that lets factory workers train robots directly on the job, without waiting for a specialist programmer to update the code.
The startup was founded in April 2025 by Jonathan Scholz, who spent seven years leading Google DeepMind’s Applied Robotics team in London, along with fellow DeepMind alumni Oleg Sushkov, Akhil Raju and Misha Denil. Their new venture applies that robotics research background to a persistent factory problem: most manual tasks are too varied, or change too often, to justify the cost of traditional fixed automation.
How the training works
Reimagine calls its approach “monkey-see, monkey-do,” a form of imitation learning: a worker demonstrates a task, watches the robot attempt it, and corrects mistakes in real time as they happen. According to the company, the robot folds each correction into its behavior immediately, without a programmer rewriting code or rebuilding a training pipeline.
“A robot should arrive with the attitude of a new colleague: ‘How can I help?’” Scholz said in the company’s launch announcement, describing the goal of letting non-specialists reshape what a robot does as work changes.
Reimagine says the approach has cut the time needed to prototype and test a new robot behavior from about a day to roughly ten minutes — fast enough, it argues, for a robot to take part in process-design conversations rather than wait on a separate engineering cycle.
Early deployments
The company has already put its physical AI systems to work in two real facilities: a made-to-order plastics business, where robots remove 3D-printer beds, operate latches and press machine controls, and a three-robot cell that recovers materials from disassembled hard drives.
Reimagine’s first phase was backed by pre-seed funding from Fly Ventures, firstminute capital and angel investors, with the amount undisclosed. The startup is now raising a new round to expand its engineering and deployment teams and put robots into more workplaces.
The launch adds to a wave of robot-learning ventures spun out of major AI labs — Google DeepMind itself unveiled Gemini Robotics 2 for whole-body humanoid control earlier this year — as founders race to move AI’s recent gains in language and reasoning into physical, hands-on work.