Maxinsights raises $30 million seed round to build the data layer for physical AI

Maxinsights has raised $30 million in seed funding led by Peak XV to turn real-world human work into robot-ready training data.

Santa Clara-based Maxinsights has raised $30 million in seed funding led by Peak XV Partners, with participation from Tola Capital, South Park Commons, INCE Capital and Alumni Ventures. The round backs a simple thesis: robots are starved not of models, but of good training data about everyday physical work.

Maxinsights builds what it calls the data layer for physical AI. More than 5,000 contractors in six countries wear its camera hardware while doing real jobs in homes, restaurants and hotels, recording first-person footage of tasks like loading a dishwasher, making a smoothie or washing a car. The company says it has captured more than 2 million hours of footage, with 1.5 million hours classified as model-ready. Unlike staged lab demos, the footage is recorded at natural speed in real settings, so models see the messy, unplanned moments that real deployments contain.

The pipeline runs in five steps: collect, check, annotate, curate and deliver. Its proprietary models label each clip across environment, task, subtask and instruction, plus hand pose, body motion and object interactions. Alongside the round, the company launched MaxLoop, an agent-in-the-loop system that scores and corrects footage as it arrives, cutting its human review team by 90%, and MaxGlove, a sensor glove developed with Genesis AI that records grip strength and object weight, the touch and force signals a camera cannot see.

The customer validation is unusually concrete for a seed-stage company. Maxinsights says it serves more than 30 customers, including frontier AI labs, with eight on recurring contracts. All 200,000 hours of egocentric data used to train Genesis AI's GENE-26.5 model came from Maxinsights, and the company says it is the largest data provider behind Dyna Robotics' Dyna-2 world-action model. It is now targeting 1 million hours of force data in 2027.

Founder and chief executive Yuguang Yuan previously co-founded Mega Gear Robotics, an industrial robotics company that was acquired. Co-founder and chief technology officer Kun Wang was an early engineer at Google X and Waymo and co-founded QCraft, whose autonomous-driving technology deployed in more than 1 million vehicles. The broader team includes engineers from Google, Microsoft, Meta, Waymo and Hulu. If physical AI funding keeps outrunning usable data, the company betting on data quality rather than data volume alone is positioned where the bottleneck is.