How data collection works
Companies are developing robots for everyday household tasks, but they lack high-quality training data. Startups including Mecka AI, XDOF and MicroAGI are working to address the problem. Bengaluru-based Humyn Labs competes with them using its own equipment with two cameras. It records people carrying out everyday tasks, then the company’s software processes and labels the footage, noting what happens in each clip.
Data is collected in more than 20 countries across Latin America and Asia. The company’s main hubs are in Brazil, India and South Korea, co-founder Manish Agarwal told TechCrunch. He said the software also tracks hand movements, interactions with objects and other details, turning raw recordings into data suitable for training AI models.
“Scale, diversity and quality matter, and they need to come together,” Agarwal said.
Founders and customers
Agarwal, the former CEO of Indian gaming company Nazara Technologies, founded Humyn Labs in February with Ishan Gupta, a former executive at brewing company AB InBev. Before that, they founded Kratos Studios, now called KGeN. KGeN has a network of members in emerging markets; the founders told TechCrunch that it helps Humyn Labs find people and collect data in their homes.
Humyn Labs has 12 paying customers, including frontier AI labs, robotics companies and world-model developers. Gupta said about 90% of revenue comes from the US but declined to name the customers.
The company generated $10 million in revenue from April to October. It has another $20 million in signed contracts, Agarwal said, while additional deals worth $20 million have yet to close.
Why the company is using venture debt
Of the $47 million raised, $26 million is equity. Accel led the round, which also included Prosus Ventures, BeeGlobal, Finserv Intelligence and Bajaj Finance’s Bajaj Alts AI Opportunities Fund, Dallas Venture Capital, Lighthouse Canton and Jump Trading.
The remaining $21 million is venture debt led by Lighthouse Canton Growth Debt Fund II, with participation from Trifecta Capital. Agarwal said the company needs working capital: it pays employees and data-collection partners daily or weekly, while AI lab customers typically pay within 30–60 days. The loan helps cover that gap. Agarwal said this form of financing dilutes founders’ and investors’ ownership less than issuing new shares.
Consent and oversight during recording
In recent months, India has become one of the hubs for collecting data for what is known as physical AI. Startups Human Archive and BuildAI also record people going about their everyday activities. The practice raises questions about privacy and consent, especially when filming takes place in people’s homes.
Agarwal said Humyn Labs records people while they do their usual work or household chores, rather than asking them to perform prepared actions. The company’s local partners help choose data-collection sites and oversee the process. Participants are paid for their work, Agarwal said, but their actions need to be observed to obtain the required data.
Humyn Labs has 51 full-time employees, most of them in Bengaluru. The company also has a network of local partners in the countries where it collects data.
Dataist’s take: Humyn Labs’ growth depends not only on the volume of recordings but also on oversight of how they are obtained. The company works in people’s homes, where privacy and consent are already central concerns. Its funding and signed contracts point to demand for this data, but some of the $20 million in deals have yet to close.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X