Maxinsights raises $30M seed led by Peak XV to sell robots 2M hours of human video

Maxinsights, a Santa Clara company that turns video of ordinary people doing chores into training data for robots, said on Oct 22 that it has raised a $30 million seed round led by Peak XV. Tola Capital, South Park Commons, INCE Capital and Alumni Ventures also invested, the company said in a press release. It launched two products alongside the round: MaxLoop, which scores footage for quality as it arrives, and MaxGlove, a sensor glove that records grip force and object weight.

The business model is labour, cameras and software. More than 5,000 contractors in six countries wear Maxinsights' own head-mounted cameras while they cook, clean, wash cars or work in restaurants and hotels; the company hires them through staffing firms and manages them directly. It says it has captured more than 2 million hours, of which 1.5 million are ready for model training, and serves more than 30 customers, eight of them on recurring contracts. Two named customers matter: all 200,000 hours of first-person data used to train Genesis AI's GENE-26.5 came from Maxinsights, and it is the largest data provider behind Dyna Robotics' Dyna-2 world-action model.

MaxLoop is the cost story. Its hand-tracking models check visibility, task completion and synchronisation on every clip, which the company says cut review costs by about 80% and its human review team by 90%, gave operators same-day feedback instead of one to two weeks, and lets it process 450,000 hours a month. MaxGlove, developed with Genesis AI, targets what video cannot show: how hard a hand grips and how contact shifts. Maxinsights says gloves on the market need long calibration and yield 10 to 20 minutes of usable capture a day, and it aims to collect 1 million hours of force data in 2027.

In MEKAWIRE's database the round matches the $30 million pre-seed that Enact, a Zurich start-up working on touch and force for robot dexterity, announced the same day, and it is half the $60 million Series B that Mecka raised on Oct 7. Alumni Ventures, one of Maxinsights' backers, also invested in Formic's 2024 round. The pattern is that money is moving down the stack from robot models to the data that trains them. MEKAWIRE's VLA explainer explains why that data is scarce.

What remains open: the release does not give revenue, and the country count differs within it (six countries for contractors, eight in the company description). Human video also has to be translated into robot actions, which is the model developers' problem and the reason force data from MaxGlove may matter more than another million hours of footage. The 2027 force-data target is the first milestone to check.