Mecka AI has raised $60 million as it works to train robots using data gathered from human movement, including information captured by body sensors and iPhones. The New York startup said the money came through two previously undisclosed rounds, a $25 million Series A that closed in November and a $35 million follow-on investment.
Framework Ventures led both financings. Other investors included Menlo Ventures, SV Angel, Kindred Ventures and Ted Xiao, an angel investor who previously worked at Google DeepMind and later joined Jeff Bezos’s AI startup Project Prometheus. Mecka AI did not disclose its valuation.
The company is betting that robotics training will increasingly depend on large-scale datasets drawn from real human activity. Its founders believe that capturing gestures, movement patterns and other physical behavior can help robots learn how to operate in the real world more effectively than approaches that rely mainly on humans directly controlling machines.
Mecka AI was founded by four entrepreneurs who came from outside the usual robotics and AI research pipeline. Josh Gao and cofounder Mogen Cheng previously built and sold a payments startup focused on restaurants. Jason Chong sold a prior company to Coinbase, while Duy Nguyen, whom Gao described as an experienced sneaker trader, made money reselling shoes.
The group shifted away from fintech after deciding that robotics represented a bigger opportunity. Gao said the team spent months reading research, visiting robotics labs and reaching out to experts before settling on an idea that had mostly lived in academic papers: training robots on human data rather than on data generated through teleoperation.
Teleoperation, in which people manually control robots, is one common method for collecting training data. Mecka AI argues that human sensor activity offers a different and potentially more scalable source. Gao said the company used custom hardware, including body sensors, along with iPhones to gather data.
The startup says it has already accumulated thousands of gigabytes of information and has signed contracts that point to an annual revenue run rate of $100 million. Gao did not name the customers behind those agreements.
Vance Spencer, a cofounder of Framework Ventures, said the firm sees Mecka AI as one of its fastest-growing portfolio companies and backed the startup because of what it viewed as strong business traction.
Mecka AI now has 40 employees and is focused on more than just collecting and organizing data. Gao said the company wants to help customers integrate robotics models and train them for deployment. He argued that useful robots can already be built and put to work now.
The startup’s approach reflects growing investor interest in so-called embodied AI, a field centered on teaching machines to act in physical environments. Companies in adjacent areas have used visual data from car cameras or other real-world recordings to improve AI systems, and Mecka AI is positioning itself as part of that broader wave.
Gao made the case that robotics is moving from research toward commercial use. As he sees it, the question is no longer whether robots can learn from real-world human behavior, but how quickly companies can turn that data into systems that are ready for deployment.