Mowito, a startup building AI foundation models for industrial robotic arms, announced on Tuesday that it has closed a $3 million pre-seed round. Version One Ventures led the raise, with All In Capital, Unisol, and iSeed joining as institutional backers. Angel investors such as AI researcher Soumith Chintala, along with founders from Foundry Robotics, Coformer.ai, and Better Capital also participated.

Mowito is taking on a problem that has persisted in manufacturing automation: even as robotic hardware has become cheaper and more capable, programming that hardware remains slow and hairy.

A single change to a product line can force engineers to spend days rewriting control code. Mowito flips that model, letting robots learn tasks by watching a human demonstrate them rather than being explicitly coded.

"Manufacturing has reached a point where hardware is no longer the bottleneck — software is. Factory robots shouldn't need to be reprogrammed every time production changes. We believe robots should learn the same way people do: by observing and repeating," said cofounder and CEO Puru Rastogi.

The startup will use the new funds to expand internationally and deepen deployments across manufacturing sites, he added.

Founded in 2024 by Rastogi, Adityanag Nagesh, and Safar V, Mowito builds software designed to run on standard, unmodified industrial robotic arms. It currently operates across Bengaluru and Detroit, supporting customers in automotive and electronics manufacturing. The new capital will fund Mowito's expansion in the United States, growth of its engineering and go-to-market teams, and continued deployment across automotive and electronics manufacturing clients.

Physical AI systems are trained to perceive and act in the physical world, rather than just process text, images, or data. Where large language models (LLMs) learn from text and generate text back, physical AI models learn from sensor data, video, and physical demonstrations, and output real-world actions.

The space has seen rising interest from investors and chipmakers alike. Manufacturing, which has structured environments and high-value repetitive tasks, is seen as one of the first places it will be tested.