Reimagine Robotics, a startup founded by four former Google DeepMind engineers, emerged from stealth on August 3, 2026. Based in London and Sydney, the company builds robots that learn new tasks directly from workers on the factory floor — without programmers. It targets factories with variable processes, where conventional automation is too expensive and too rigid.
Key takeaways
- Reimagine Robotics left stealth on August 3, 2026, and was founded in April 2025.
- The four founders come from Google DeepMind's Applied Robotics team.
- CEO Jonathan Scholz led that team for seven years.
- Pre-seed funding comes from Fly Ventures, firstminute capital and angel investors.
- On a hard drive recycling project, prototype build time dropped from about one day to roughly 10 minutes.
Four engineers from DeepMind
Behind Reimagine Robotics are Jonathan Scholz, Oleg Sushkov, Akhil Raju and Misha Denil. All four previously worked on the Applied Robotics team at Google DeepMind. Scholz founded that team in London and led it for seven years.
The company was created in April 2025 and is only now going public. It keeps two headquarters — London and Sydney. The pre-seed?pre-seed: The earliest startup funding round, preceding the main seed round. round was backed by Fly Ventures, firstminute capital and undisclosed angel investors. The startup is already seeking further capital to grow its team and add deployments.
Robots that learn by watching
Reimagine’s idea is simple. Instead of writing code for every task, a worker shows the robot what to do and corrects it when it makes a mistake. Scholz calls it a "monkey-see, monkey-do" approach. Process knowledge comes from the operator, not a specialist integrator.
A useful robot should be able to learn from the person doing the work. They should be able to show it a task, put it right when it makes a mistake, and move on to the next problem.
Jonathan Scholz, co-founder and CEO of Reimagine Robotics.
It reverses the classic model, in which a specialist engineer designs the robot integration. Here the robot adapts to the human, not the other way around — and a new task no longer takes weeks of a programmer’s work.
Where the robots already work
The company showed two deployments. At a plastics manufacturer, robots tend 3D printers overnight — removing print beds, operating latches and pressing control buttons. Workers later extended the automation to washing, curing and drying the prints.
The second deployment is a three-robot cell for dismantling hard drives and recovering materials. That is where the time to prepare a working prototype fell from about one day to roughly 10 minutes.
A niche the giants avoid
Reimagine targets plants with variable processes and volumes too small to justify conventional automation. That is a different market from the one general-purpose humanoid makers such as Figure or 1X are chasing with a single universal robot for many tasks.
The demonstration-based learning approach echoes the direction of robotic VLA models. The difference is that Reimagine sells not a humanoid but a method for quickly deploying ordinary robotic arms to a specific, narrow job.
Why it matters
The biggest barrier to factory automation is not the hardware itself but the cost and time of programming it. If a robot learns from an operator in minutes, the economics of automation shift toward small and mid-sized plants long cut off from the technology. The founders' DeepMind pedigree adds credibility, but the company is still at pre-seed and has shown only two deployments. Its significance will hinge on whether the method scales beyond carefully chosen demonstration cases.
What's next?
- The company is seeking a further funding round to expand its team and add deployments — by its own account.
- Reimagine has not yet disclosed a product name or target hardware platform — key information for judging scalability.
- Open question: whether demonstration-based learning holds up on tasks more complex than tending 3D printers and dismantling drives.
