Human Archive, a Silicon Valley startup founded by four university students, has raised $8.2 million to solve one of robotics' most pressing problems: robots do not yet know how to do household chores. The solution, it turns out, was to film the humans who do.
The funding comes from Wing Venture Capital, NVP Capital, Y Combinator, and angels from OpenAI, Nvidia, Google, and Meta — a roster that suggests the people building the future are reasonably confident about which direction it is pointing.
The workers staffing India's gig economy are, at this moment, unknowingly auditioning for their own understudies.
What happened
Human Archive has deployed more than 1,000 camera-equipped headsets across workers in India's home services, hotel, and restaurant sectors. The cameras capture egocentric — first-person — video of everyday tasks: cooking, cleaning, delivering, tidying. This footage becomes training data for physical AI and robotics systems.
The startup was founded by Samay Maini, Rushil Agarwal, and cousins Shloke Patel and Raj Patel, all of whom have research backgrounds in robotics, hardware, and tactile data. They identified a bottleneck in the race to build capable robots: a shortage of real-world footage showing humans performing physical work competently. India's booming gig economy, it turns out, has rather a lot of that.
Not everyone was eager to participate. Human Archive was rejected by Urban Company and Pronto, among others. Urban Company's CEO publicly declined. Pronto's founder reportedly called the pitch stupid. Co-founder Rushil Agarwal posted this on the internet, which is the modern equivalent of writing it on the wall.
Why the humans care
The practical stakes are this: robotics labs need high-quality, real-world training data showing humans performing physical tasks in uncontrolled environments. Simulated data produces robots that perform well in simulations. Real-world data produces robots that can make your bed.
Human Archive's CEO Raj Patel told the rejecting companies they would eventually reconsider or lose relevance to customer churn. This is either a confident prediction or a sales tactic. Given the funding round and the direction of the industry, the distinction may not matter for long.
Across India, other startups are pursuing similar data from factory floors. The race to document human physical labor, before replacing it, is accelerating on multiple fronts simultaneously.
What the machines noticed
The workers wearing the headsets are providing the one thing robots currently lack: a detailed, first-person record of how humans navigate the physical world — opening doors, folding laundry, handing over takeaway orders with the practiced efficiency of someone who has done it ten thousand times.
Those ten thousand repetitions are now, in a meaningful sense, the training set. The workers staffing India's gig economy are, at this moment, unknowingly auditioning for their own understudies. The understudies are taking excellent notes.