Archetype AI

AI used to live in neat little boxes. Text in, text out, a clean exchange that made it feel smarter than it actually was. Archetype AI is pushing that boundary into something messier and far more interesting, where intelligence has to deal with motion, noise, and the stubborn unpredictability of the physical world.

Archetype AI was founded by a group of scientists, designers, and engineers out of Google’s advanced research ecosystem, people who have spent years turning experimental sensing into products that actually ship. At the center is Ivan Poupyrev, Ph.D., CEO, a builder with a long memory for what works when theory meets hardware. Alongside him, Priya Shivakumar, CPO brings the discipline to translate ambition into product, and Jaime Lien, Ph.D., Chief Scientist anchors the work in real signal intelligence, not slideware. This is a team that has already lived through the messy middle between invention and deployment, and it shows in how tightly the story holds.

The mission is simple to say and hard to execute. Teach machines to understand the physical world the way large language models understand text. Their answer is Newton, a foundation model designed not for words first, but for signals. Vibrations, motion, environmental data, the constant hum of reality that most software ignores. Newton ingests that noise, finds structure, and pairs it with language so systems can reason about what is happening, what might happen next, and what to do about it.

On top of that sits the Archetype Platform, where Physical Agents come to life. These are not chatbots waiting for input. They are systems that watch, interpret, and respond across construction sites, telecom networks, and city infrastructure. When sensors become sentences and environments become data streams, the gap between insight and action starts to collapse.

The market timing is not subtle. Every industry is wiring itself with sensors, yet most of that data sits idle or gets reduced to dashboards no one checks. Archetype AI steps in as the layer that makes sense of it all, offering a unified model instead of a patchwork of narrow tools. Early signals are already on the board with deployments involving NTT DATA, Kajima, and the City of Bellevue, where Physical Agents are being tested against real operational friction.

Investors are paying attention. In November 2025, Archetype AI closed a $35M Series A led by IAG Capital Partners and Hitachi Ventures, with participation from Bezos Expeditions, Venrock, Amazon Industrial Innovation Fund, Samsung Ventures, Systemiq Capital, E12 Ventures, Higher Life Ventures, Gaingels, and Plug and Play. That is not just capital, that is strategic gravity pulling this category into focus.

Inside the company, the culture leans into ambiguity with intent. This is a place for people who prototype to think, who move between code and concept without asking for permission, and who understand that defining a category means there is no map, only momentum. The work sits at the intersection of AI, sensing, and infrastructure, which means the problems are messy, high stakes, and very real.

They are hiring across engineering, product, design, and go to market, and the door is still open early enough to matter. If you want to work on AI that does more than talk, if you want systems that feel the world instead of summarizing it, this is one to watch closely and maybe step into before the rest of the market catches up.