Startups & Funding

Humyn Labs raises $47M for real-world robot training data

Humyn Labs raised $26 million in equity and $21 million in venture debt to expand two-camera collection and labeling of real-world robotics data across more than 20 countries.

India’s Humyn Labs raises $47M to feed U.S. robotics companies the data they need
India’s Humyn Labs raises $47M to feed U.S. robotics companies the data they needgeralt / pixabay
By Rebecca Stone5 min read

Why it matters

  • Series A financing totals $47 million: $26 million in equity and $21 million in venture debt.
  • Humyn Labs collects data in more than 20 countries, with Brazil, India and South Korea as main hubs.
  • The startup generated $10 million in revenue from April through October, with $20 million signed and $20 million pending.
  • About 90% of revenue comes from the United States, according to co-founder Ishank Gupta.
  • Humyn Labs has 12 paying customers and 51 full-time employees.

By Rebecca Stone

Humyn Labs has raised $47 million in Series A financing to expand a real-world data pipeline for training robots to perform everyday tasks. The Bengaluru startup says U.S. AI and robotics companies need more high-quality examples of human activity as they build models that operate in physical environments.

The financing consists of $26 million in equity and $21 million in venture debt. The capital backs capture and labeling work across Latin America and Asia, where Humyn Labs already collects data in more than 20 countries. The deal targets a constraint in physical AI: developers still lack enough usable training data from human settings.

Why is real-world data the bottleneck?

Robotics companies around the world are trying to build machines that can perform everyday human tasks. Humyn Labs says obtaining enough quality training data remains difficult. The company competes with startups including Mecka AI, XDOF and MicroAGI, which also collect real-world data for robotics training.

The shortage supports a separate market for collection and annotation services. Humyn Labs says its customers include frontier AI labs, robotics companies and developers of world models. Its service combines proprietary capture hardware with software processing and a network of local partners.

That places Humyn Labs in the data supply chain rather than the finished-robot market. Its customers use the recordings to train systems that operate in physical environments. Its operating model pairs geographic reach with detailed annotation.

Who backed the $47 million financing?

  • Accel led the $26 million equity portion alongside Cactus Partners. Prosus Ventures, BeeGlobal, Bajaj Finance’s Finserv Intelligence, Bajaj Alts AI Opportunities Fund, Dallas Venture Capital, Lighthouse Canton and Jump Trading also participated.
  • Lighthouse Canton Growth Debt Fund II led the $21 million venture-debt portion. Trifecta Capital also participated.

The split gives Humyn Labs both expansion capital and a source of working capital. The venture-debt component addresses a payment gap between company outlays and customer receipts, Agarwal said.

How does Humyn Labs turn daily activity into training data?

Humyn Labs developed its own two-camera hardware to record people performing everyday tasks. Its software processes and labels the footage, tagging what happens in each clip so an AI model can learn from the recording.

The company also uses software to track hand movements, interactions with objects and other details. That processing turns raw recordings into structured training data. “Scale, diversity, and quality are important aspects, and they need to work in tandem,” Agarwal said.

Humyn Labs collects data in more than 20 countries across Latin America and Asia. Brazil, India and South Korea serve as its main hubs, co-founder Manish Agarwal said. The company combines that broad footprint with its two-camera system and labeling software.

Agarwal previously served as CEO of Indian gaming company Nazara Technologies. He founded Humyn Labs in February with Ishank Gupta, a former executive at AB InBev. The two previously founded Kratos Studios, now known as KGeN.

KGeN has a network of contributors across emerging markets, the founders told TechCrunch. That network helps Humyn Labs collect data in people’s homes, extending the startup beyond workplaces and formal collection sites.

What traction has Humyn Labs reached?

Humyn Labs says it has 12 paying customers. They include frontier AI labs, robotics companies and developers of world models. About 90% of the company’s revenue comes from the United States, Gupta said, although he declined to name the customers.

About 90% of revenue comes from the United States even though the collection network spans Latin America and Asia. The gap between where data is recorded and where revenue comes from defines the startup’s cross-border operating model.

The startup generated $10 million in revenue between April and October. It has another $20 million in signed contracts and a further $20 million in deals that have not been finalized, Agarwal told TechCrunch.

The figures place the financing against revenue already generated, signed contracts and un finalized deals. That mix makes working-capital timing material.

Cash timing has become a material constraint. Humyn Labs pays workers and data collection partners daily or weekly, but it typically waits 30 to 60 days to receive payment from AI lab customers. Agarwal said the growth in the business increased its need for working capital.

The $21 million debt component bridges that gap. Borrowing allows Humyn Labs to keep paying collection workers and partners without waiting for customer checks. Agarwal said borrowing is less dilutive than selling more equity.

How does India address privacy and consent?

Over the last few months, India has emerged as a hub for collecting training data for so-called physical AI, TechCrunch reported. Human Archive and BuildAI have tapped the country’s large workforce to record people performing everyday tasks.

Both Indian competitors use labor as the collection mechanism: they tap India’s large workforce to record routine tasks. Humyn Labs adds dedicated two-camera hardware and its own labeling software to that collection model.

The practice has raised questions about privacy and consent, particularly when cameras record people inside their homes. The concern is sharper for home recordings because collection takes place in private settings and can include interactions that software later labels.

Agarwal said Humyn Labs records people doing their usual work or household chores rather than asking them to perform staged activities. The startup works with local partners to identify collection sites and oversee the process. Local partners oversee collection at those sites.

“The worker is a worker who is incentivized, but he needs to be monitored and supervised for the right input,” he stated. The comment sets out Humyn Labs’ approach to collection: incentivized participation, local oversight and a focus on natural activity rather than scripted behavior.

How large is the operation?

Humyn Labs currently has 51 full-time employees, most of them based in Bengaluru. It also relies on a network of local partners across the countries where it collects data, Agarwal said. The 51 employees sit alongside a partner network that extends across the collection footprint.

The next test is conversion: Humyn Labs must turn $20 million in pending deals into revenue while preserving the scale, diversity and quality Agarwal says must work together.

Source: techcrunch.com

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

728 articles

Related articles

  1. micro1 Pays Upwards of $2 Million for Operational AI Data
  2. US Data Brokers Sell $500 Million in RL Environments to Chinese AI Labs
  3. Anthropic maps $518bn infrastructure spend as IPO backers eye $2tn valuation
  4. Eka Robotics Raises $150M at $1.45B Valuation on Force-First Robot AI
  5. Startups Will Pay You to Rent Your Idle GPU for AI Inference

« Previous articleNext article »