Data is the oil of the AI boom—and the startup Mecka AI hopes to tap the vast reservoir of hand gestures, walking gaits, and immense collection of humanity’s physical motions to better train our future overlords: robots.

So-called “embodied” AI data isn’t exactly novel. Wayve, the British autonomous driving company, has trained its models on visual input pulled from car cameras. And, on Friday, the startup MicroAGI announced that it was offering New Yorkers a free home cleaning—if they consented to cameras recording the cleaners’ every action.

Still, Josh Gao, Mecka AI’s cofounder and CEO, believes that his startup is a first mover—and his investors believe he has a chance to strike it big. Headquartered in New York City, the company has raised a total of $60 million across two previously unannounced fundraises: a $25 million Series A round closed in November and $35 million in a follow-on investment. Gao declined to tell me at what valuation he raised the capital.

Framework Ventures, a VC best known for its crypto investments, led the rounds. Other investors who participated included Menlo Ventures, SV Angel, Kindred Ventures, and the angel Ted Xiao, a former researcher at Google DeepMind and founding member at Jeff Bezos’s AI startup Project Prometheus.

“Many folks are now looking and realizing that robotics is going to be one of the most important waves that’s happening,” Gao said on a video call from the Chinese city of Shenzhen, where he was visiting a factory that was producing custom equipment his startup designed to capture human data.

Gao and his three cofounders aren’t your conventional AI hotshots with credentials from leading AI labs or Stanford University. In 2023, Gao and Mecka AI cofounder Mogen Cheng sold their prior startup, which focused on building out payments tech for restaurants. Their third cofounder, Jason Chong, sold his own startup to the crypto exchange Coinbase, and their fourth, Duy Nguyen, was an “OG sneakerhead” who made millions of dollars “flipping” shoes, said Gao.

Rather than focus on fintech, the founding quartet, who knew each other through mutual friends, decided to look at where the puck was going and zeroed in on AI for robots. “Can you have general intelligence in the physical world?” asked Gao.

They spent months devouring research papers, visiting robotics labs, and messaging roboticists on social media. Soon, they landed on what Gao said was an unconventional hypothesis: Train robots on data collected from humans, not from data collected through “teleoperation,” or when humans manually operate a robot.

Read More: Mecka AI raises $60 million to train robots with human data