# Indian Workers Earn $2.62 an Hour Filming Themselves to Train Robots > Source: > Published: 2026-09-09 02:50:02+00:00 *Tens of thousands of Indian workers are strapping cameras to their own heads and filming themselves at work, hour after hour, so humanoid robots can learn to take their jobs.* Bloomberg reported on September 8 that gig workers across India are wearing head-mounted cameras, often just an iPhone velcroed to a headband, to capture what the robotics industry calls "egocentric data": the precise hand angles, grip pressure, and sequencing behind tasks like stripping labels off plastic bags, stitching shoes, welding steel, and sorting packages on a warehouse floor. The pay works out to roughly $2.62 an hour. The buyers are the humanoid robot makers racing to get machines out of demo videos and onto factory floors. This isn't a one-off story. Bloomberg first reported on the pipeline in August, describing thousands of Indian workers filming their own labor for AI firms building the robots meant to replace them. The September piece confirms the trade has only grown since, with more startups, more workers, and more buyers lining up for the footage. Here's how the pipeline actually works. A camera, angled down toward the hands, records a worker performing a task they already know how to do. That footage gets processed into a dataset and sold to robotics companies, who feed it into models that teach a robot's arm and gripper how a human hand actually moves through the world. According to Bloomberg's reporting, the end customers for this kind of data include Figure AI, Agility Robotics, 1X Technologies, Boston Dynamics, and Chinese firms Unitree and UBTECH. None of them film the factory floor themselves. They buy the footage instead. One of the companies driving this is Egolab.AI, founded in January by Raghav Samani, 19, and Varun Pareek, 18, out of Maharashtra. The startup describes itself as India's largest first-person point-of-view data aggregator, and it moved fast: by March, it had already been bought by Build Artificial Intelligence Inc., a Delaware-registered firm whose founders live in San Francisco while its India operations run out of Bengaluru. Egolab has said its footage ends up with companies including Tesla, Boston Dynamics, and Figure AI, according to company documents cited in earlier Bloomberg coverage. It's a strange kind of pipeline. A startup barely out of its teens, packaging the muscle memory of Indian factory and warehouse workers, and selling it to the best-funded robotics labs in the world. [BLS Widens Its AI Job Risk List to Include PR, Web Design and Hotel Work](https://startupfortune.com/bls-widens-its-ai-job-risk-list-to-include-pr-web-design-and-hotel-work/) Bloomberg reported on September 1, 2026 that the BLS expanded its AI exposure categories to include PR specialists, web developers and hotel workers, broadening an official list that already covered paralegals, graphic designers and customer service roles. The expansion follows Bloomberg's July reporting that tech and finance sectors are losing... - [which jobs will be replaced by artificial intelligence](https://startupfortune.com/bls-widens-its-ai-job-risk-list-to-include-pr-web-design-and-hotel-work/) - [AI job displacement risk for service industry workers](https://startupfortune.com/bls-widens-its-ai-job-risk-list-to-include-pr-web-design-and-hotel-work/) Objectways, another firm working the same market, pays around 250 rupees, about $3, per hour for footage of workers doing household chores: chopping vegetables, folding laundry, wiping counters. That's the domestic-robot end of the same trade. The industrial end runs through steel plants and shoe factories. Either way, the arithmetic is the same: cheap, plentiful footage of real human motion, priced at a fraction of what it would cost to shoot and label in the US. There's a version of this labor market that's already familiar. Time reported in 2023 that Kenyan contractors were paid about $2 an hour to label the toxic content that trained OpenAI's safety filters. The India robotics pipeline is a cousin of that trade, not a copy of it. RLHF workers filtered a chatbot's outputs. These workers are handing over the physical technique of their own jobs, frame by frame, to a machine designed to eventually do that job without them. Many of the workers Bloomberg spoke to said they weren't told that the footage would be used to automate the exact job category they were filming themselves doing. That's the detail that sticks. Nobody is being tricked into unemployment on purpose, most likely, but nobody is volunteering for it either. ## Why this matters for the robots everyone's betting on Humanoid robotics is one of the most heavily funded corners of the AI trade right now. Figure has raised money from Nvidia and Microsoft-adjacent backers. Tesla keeps pushing timelines for Optimus. Morgan Stanley has forecast that more than a billion humanoid robots could be in use globally by 2050. Every one of those forecasts assumes the hard part, teaching a robot to move a hand the way a human does, gets solved somehow. This is the somehow. It isn't a breakthrough in a lab. It's tens of thousands of hours of footage from garment workers in Maharashtra and steel workers wherever the next contract lands, priced at a couple of dollars an hour, stitched into training sets. That's less a sign robots are already good than a sign of how much raw human data still has to be bought and processed before they are. The gap between demo-day humanoids and warehouse-ready ones is being closed the same way most AI gaps get closed: with cheap labor doing the unglamorous work nobody funds a keynote around. **Also read:** [Kioxia CEO Rules Out an SK Hynix Merger While Vowing to Tame Chip Prices](https://startupfortune.com/kioxia-ceo-rules-out-an-sk-hynix-merger-while-vowing-to-tame-chip-prices/) • [A Lawsuit Says Anthropic's Claude Max 20x Delivers Far Less Than Advertised](https://startupfortune.com/a-lawsuit-says-anthropics-claude-max-20x-delivers-far-less-than-advertised/) • [Hackers Are Draining Claude Subscribers Usage Without Stealing a Password](https://startupfortune.com/hackers-are-draining-claude-subscribers-usage-without-stealing-a-password/) [Google DeepMind's Safety Team Warns Job Seekers Its Own AI Hiring Filters Misfire](https://startupfortune.com/google-deepminds-safety-team-warns-job-seekers-its-own-ai-hiring-filters-misfire/) Bloomberg reports that Google DeepMind's AGI Safety and Alignment Team tells job applicants to submit a special workaround form because it doesn't trust Google's own AI hiring filters. The twist: Google markets that same AI screening technology to corporate clients as a selling point. - [AI hiring bias in recruiting tools](https://startupfortune.com/google-deepminds-safety-team-warns-job-seekers-its-own-ai-hiring-filters-misfire/) - [why AI filters reject qualified candidates](https://startupfortune.com/google-deepminds-safety-team-warns-job-seekers-its-own-ai-hiring-filters-misfire/) *This article is posted in [AI News](https://startupfortune.com/category/ai/), check it out for more related stories.* ## Join the discussion [Open in the community →](/community/) Almost there. Sign in and your reply posts straight away.