Maxinsights, a Santa Clara-based company building the data layer for physical AI, has raised US$30 million in Seed funding led by Peak XV. The round includes participation from Tola Capital, South Park Commons, INCE Capital, and Alumni Ventures. Alongside the funding, the company launched MaxLoop, a quality control platform, and MaxGlove, a sensor glove for tactile and force data.
Addressing the Physical AI Data Gap
Investors placed US$47.4 billion into physical AI in the first half of 2026, nearly four times the total from the second half of 2025. Capital is moving faster than the training data available for robotics models. Maxinsights aims to close this gap by recording real human activity at natural speed in homes, restaurants, and hotels.
A Full Data Pipeline
Unlike competitors that stage tasks in mock environments, Maxinsights captures real work in real settings and manages every step from capture to delivery in-house. The company has collected more than 2 million hours of footage, with 1.5 million hours model-ready and available for immediate delivery. It serves over 30 customers, including frontier AI labs and foundation model companies, with eight on recurring contracts.
A Global Collection Network
More than 5,000 contractors in six countries wear Maxinsights' proprietary camera hardware while they load dishwashers, make smoothies, or wash cars. The company hires them through staffing firms in each market and manages them directly rather than handing collection off to a vendor. This approach produces clearer footage compared with shaky iPhone or GoPro videos.
MaxLoop Automates Quality Control
MaxLoop automatically scores footage as it arrives, giving operators feedback the same day instead of within one to two weeks. The platform has reduced review costs by roughly 80% and allows Maxinsights to process 450,000 hours of video each month. The same models also extract pose, object states, contact, and force from each clip.
MaxGlove Captures Touch and Force
MaxGlove, developed with Genesis AI, records grip force, object weight, and contact shifts that cameras cannot see. Force and torque signals reveal contact, pressure, resistance, and slip, helping robots understand whether they have grasped an object securely. Maxinsights targets 1 million hours of force data in 2027 using the glove across its collection network.
Traction with Robotics Foundation Models
Two closely watched robotics foundation models in 2026 were trained partly on Maxinsights data. Genesis AI used 200,000 hours of egocentric footage provided exclusively by Maxinsights to train GENE-26.5, while Dyna Robotics listed Maxinsights as its largest data provider for Dyna-2. These collaborations show that foundation model developers are adopting real-world egocentric data as a training standard.
Leadership and Investor View
Maxinsights was founded by CEO Yuguang (YG) Yuan, who previously co-founded Mega Gear Robotics, and CTO Kun Wang, a former early engineer at Google X and Waymo. The broader team includes engineers from Google, Microsoft, Meta, Waymo, and Hulu. Peak XV Managing Director Rohit Agarwal said the company treated quality control as an engineering problem, which is difficult to copy after the fact.
Maxinsights is positioning itself as a foundational data provider for physical AI at a time when robot models require large volumes of varied real-world interaction. Its combination of proprietary hardware, automated quality control, and a global collection network supports a recurring customer base. The company's product roadmap points to an expanded role for tactile data, positioning it to shape how physical AI models learn from human work.