The robots of the future are learning to fold laundry, stack shelves and cook by watching humans, and a nine-month-old Indian startup has become one of their teachers.
Humyn Labs, a Bengaluru-based company that collects real-world data for training robotics and physical AI models, has raised $47 million in a Series A round, TechCrunch reported on Wednesday, November 18. The round comprises $26 million in equity led by Accel and $21 million in venture debt led by Lighthouse Canton Growth Debt Fund II, with participation from Trifecta Capital.
Other equity investors include Cactus Partners, Prosus Ventures, BeeGlobal, Bajaj Finance's Finserv Intelligence, Bajaj Alts AI Opportunities Fund, Dallas Venture Capital, Lighthouse Canton and Jump Trading, according to TechCrunch.
Serial founders
Humyn Labs was founded in February 2026 by Manish Agarwal, the former chief executive of listed gaming company Nazara Technologies, and Ishank Gupta, a former executive at brewer AB InBev, according to TechCrunch. The two previously founded Kratos Studios, now known as KGeN, a gaming and Web3 platform.
How it works
Humyn Labs has built its own two-camera hardware to record people performing everyday tasks, according to TechCrunch. It then processes and labels the footage so AI systems can learn from it, including tracking hand movements and how people interact with objects.
Rather than staging activities in studios, the company collects data mainly in people's homes and workplaces, working with local partners to identify sites and supervise collection. It operates in more than 20 countries across Latin America and Asia, with Brazil, India and South Korea as its main hubs.
"Scale, diversity, and quality are important aspects, and they need to work in tandem," Agarwal told TechCrunch. He added that contributors are incentivised but need to be "monitored and supervised for the right input."
Rapid revenue growth
The company has grown quickly. Humyn Labs generated $10 million in revenue between April and October and has $20 million in signed contracts, with another $20 million in deals not yet finalised, according to TechCrunch. It has 12 paying customers, including frontier AI labs, robotics companies and developers of world models, a type of AI that simulates physical environments. About 90% of its revenue comes from the United States.
The company has 51 full-time employees, most of them in Bengaluru.
Why venture debt
The large venture debt component is unusual for a Series A but reflects the company's business model. Humyn Labs pays its contributors daily or weekly but typically waits 30 to 60 days for customers to pay, according to TechCrunch. The debt funds this working capital gap, allowing the company to grow without diluting shareholders as much as an all-equity round would.
The physical AI data rush
Humyn Labs is riding one of the hottest trends in artificial intelligence. Companies building humanoid robots, household robots and industrial automation need vast amounts of data showing how humans perform physical tasks. Unlike text and images, which are abundant on the internet, high-quality data on physical manipulation is scarce.
First-person video, recorded from the perspective of the person doing the task, is particularly valuable because it shows hand movements and the sequence of actions. Robotics developers combine such data with simulation and data collected by robots themselves to train their systems.




