According to Robotics & Automation News, Maxinsights has raised $30 million in seed funding to expand a physical AI data platform that pays people to record everyday activities — cooking, cleaning and other work — to generate training data for robots and AI models.
The round was led by PeakXV, with participation from Tola Capital, South Park Commons, INCE Capital and Alumni Ventures. The company was founded by robotics and autonomous-driving veterans, and its wider engineering team includes former employees of Waymo, Google, Microsoft and Meta.
Maxinsights says it has captured more than 2 million hours of real-world human activity, of which 1.5 million hours is model-ready, and that it serves more than 30 customers, including AI labs and foundation model developers.
Contractors record real homes, not staged sets
More than 5,000 contractors across six countries use Maxinsights’ camera hardware to record themselves carrying out everyday activities, ranging from loading dishwashers and preparing food to washing cars.
Unlike staged data collection, the approach is intended to expose AI models to the variation and unexpected events encountered in real homes and workplaces. The company argues that robots are deployed into real kitchens rather than staged ones, where something unexpected happens every few minutes.
Maxinsights has also launched two supporting tools. MaxLoop is an AI-based system for monitoring and improving the quality of collected footage, while MaxGlove is a sensor glove designed to capture tactile information such as grip force and object weight that cannot be obtained from video alone.
Quality control treated as an engineering problem
MaxLoop automatically assesses incoming footage and provides feedback to operators. According to the company, the system has reduced review costs by approximately 80 percent and enables it to process 450,000 hours of footage a month.
MaxGlove is being developed with Genesis AI to capture force and touch information, including pressure, resistance and slip. Maxinsights is targeting 1 million hours of force data in 2027 using the glove and its existing collection network.
The company says all 200,000 hours of egocentric data used to train Genesis AI’s GENE-26.5 model came from Maxinsights. It also says it is the largest data provider behind Dyna Robotics’ Dyna-2 World-Action Model.
“We built the whole pipeline ourselves, from the hand-tracking models to the platform that scores every operator’s footage as it arrives, because nobody was going to sell us the pieces that make messy real-world data usable.” — Yuguang (YG) Yuan, founder and CEO of Maxinsights
Yuan added that collecting one great hour of data is easy, and that holding quality across millions of hours and thousands of people is what the company was built to do.
Investors back the quality-first approach
Rohit Agarwal, managing director of Peak XV, said almost every company in the category is selling hours, and positioned Maxinsights as the only one that treated quality control as an engineering problem rather than a staffing cost.
“That decision is hard to copy after the fact, and it’s why the labs that tried other vendors keep coming back to them.” — Rohit Agarwal, managing director of Peak XV
Johnson Wang, cofounder of Genesis AI, said robots need to learn from the complexity and variation of everyday life, and that carefully controlled datasets alone cannot capture it.
“With Genesis 26.5, we focused on footage of people performing everyday tasks naturally, with much of that data provided by Maxinsights, to help our models learn from a broader range of real-world situations.” — Johnson Wang, cofounder of Genesis AI
Yuan previously co-founded industrial robotics company Mega Gear Robotics, which was acquired. Maxinsights co-founder and CTO Kun Wang was an early engineer at Google X/Waymo and later co-founded autonomous driving company QCraft.
Source: Robotics & Automation News
Compiled from international media by the SCI.AI editorial team.
