Humyn Labs, a Bengaluru startup that records people performing ordinary tasks and sells the footage as robot training data, has raised $47 million in Series A financing. The round splits into $26 million of equity led by Accel, with Prosus Ventures, BeeGlobal, Bajaj Finserv Intelligence and the Bajaj Alts AI Opportunities Fund, Dallas Venture Capital, Lighthouse Canton and Jump Trading participating, and $21 million of venture debt led by Lighthouse Canton Growth Debt Fund II with Trifecta Capital.
The company was founded in February by Manish Agarwal, former chief executive of the Indian gaming company Nazara Technologies, and Ishank Gupta, a former AB InBev executive. The two previously built Kratos Studios, now KGeN, whose network of contributors across emerging markets gives Humyn Labs a ready pool of people to record in their own homes.
The hardware is Humyn built: a two-camera rig that captures first-person footage of someone doing their usual work or household chores rather than a staged demonstration. Software then processes the recordings, tracking hand movements and object interactions and tagging what happens in each clip so that a model can learn from it. Agarwal argues that scale, diversity and quality have to work together, which is why collection runs across more than 20 countries in Latin America and Asia, with Brazil, India and South Korea as the main hubs. Local partners identify sites and supervise the work.
The commercial numbers are unusually specific for a company this young. Humyn Labs says it has 12 paying customers, including frontier AI labs, robotics companies and developers of world models, and that about 90 percent of revenue comes from the United States. It generated $10 million of revenue between April and October, holds another $20 million in signed contracts, and has a further $20 million in deals not yet finalised. It employs 51 people full time, most in Bengaluru.
That is also why the round includes debt. Humyn Labs pays workers and collection partners daily or weekly but typically waits 30 to 60 days for payment from AI lab customers, so borrowing bridges the gap with less dilution than selling more equity. It is a working-capital problem disguised as a funding round, and it says something about the business: this is a labour and logistics operation with software attached, not a pure model company.
Competition is filling in fast. Mecka AI raised $60 million from Sequoia in October, XDOF was reported in September to be in talks for a Series B at a $1.2 billion valuation three months after leaving stealth, and MicroAGI, Human Archive and BuildAI are chasing the same buyers. India has become a hub for physical AI data collection precisely because it has a large workforce willing to be filmed at home.
That last point is the exposure. Recording people inside their homes raises consent and privacy questions that are not solved by a partner network, and Indias IT ministry has been examining how home-services platforms collect data for physical AI training. Whoever industrialises this collection fastest also inherits the regulatory scrutiny that comes with it.
The strategic read is that the bottleneck for physical AI has moved. Labs can now buy robot bodies and simulation, but they cannot buy the messy human demonstration data that makes a policy generalise, and they are outsourcing that to companies that run field operations. Humyn Labs has roughly $30 million of contracted revenue at eight months old, which is a real business. Whether it is a durable one depends on whether its label quality stays ahead of cheaper collectors and whether consent practices survive contact with regulators in 20 countries.
Comments (0)
Log in to join the discussion
Log InNo comments yet