Munich humanoid startup Vesoma exits stealth with 60-person team, DeepMind AI chief
What's the deal? VesomaDealroom has a profile for this one. Try Dealroom →, the European startup behind the Munich-based Kyle humanoid prototype, has emerged from stealth with a team of more than 60 people, a first product called Vesoma 1, and a bet that robots can learn physical work through their own interactions. The company, which began work in December 2025, published its introduction on September 22 and said the day after that the first Vesoma 1 units are being built.
Who's behind it? Nikolai Ensslen, co-founder of motion-control specialist Synapticon, is chief executive officer and co-founder. Fellow co-founder Peter Skoromnyi, who co-founded games company EasybrainDealroom has a profile for this one. Try Dealroom →, is chief strategy officer, pairing industrial robotics experience with consumer software know-how.
The DeepMind link: Martin Riedmiller, a former research director and Control Team lead at Google DeepMind, serves as chief AI officer. Vesoma's design head Marcelo Gutierrez previously worked on TeslaDealroom has a profile for this one. Try Dealroom →'s Optimus, Apptronik's Apollo, and CartwheelDealroom has a profile for this one. Try Dealroom →'s Yogi.
What's the endgame? Vesoma argues demonstrations alone will not deliver the reliability customers need. It wants an agent that improves through its own interactions with the physical world, backed by hardware built around that learning process. It targets manufacturing, logistics, and warehousing first, and rules out weapons, military, and surveillance uses.
Why now? The startup says it designed, built, and walked its first prototypes within six months using a learned locomotion policy, and has fitted out more than 3,500m² of lab and office space in Munich, Germany, and Limassol, Cyprus. The public launch adds detail on leadership and approach that was missing when the Kyle prototype first surfaced.
What could go wrong? The polished robot images in Vesoma's media kit are labelled design visualisations, not photographs of a finished machine. The company has published no manipulation success rates, intervention figures, or full-shift results to prove its learning approach beats systems trained on human demonstrations.
The signal: Vesoma joins a growing cohort betting that Europe's industrial expertise can support a competitive humanoid business. Whether learning through interaction outperforms imitation remains a thesis, not a verdict — and the next test will be moving from walking demos to reliable, contact-heavy work.
Read more: humanoidsdaily.com
Image credit: Vesoma
