The Silicon Valley-based startup Human Archive is responding to this trend by forming partnerships with companies to have workers wear special camera-equipped hats and collect video data from a first-person perspective for training robotics. Without naming specific partners, the company said it works with players in the home services, dormitories, and restaurants sectors to collect egocentric data, and has more than 1,000 active headsets deployed in several locations.
Against this backdrop, Human Archive announced raising $8.2 million in funding from Wing Venture Capital, NVP Capital, Y Combinator, and angels from OpenAI, Nvidia, Google, Mercor, AfterQuery, BAIR, SAIL, Brad Boa, and Meta.
Team, Technology Base, and Regulatory Challenges
The company was founded by four students from Berkeley and Stanford – Samay Mani, Rushil Agarwal, Shalek Patel, and Raj Patel; the latter serves as chief executive officer. All four have academic backgrounds in robotics, hardware, and tactile data.
The founding of the company was a forecast of where the AI industry is headed. At a time when robotics laboratories and leading AI companies strive to create machines that can perform physical tasks in the real world, they face a shortage of high-quality training data. Human Archive believes that workers in India’s gig economy constitute an untapped, scalable resource for precisely such data. Despite active partnerships, the company admitted that many Indian players in home services declined to cooperate, including Pronto and Urban Company.
Data Technologies and Market Engagement
It is evident that the reluctance of major players became a public talking point: Entrackr reported that Pronto is actively seeking partners to collect workers’ data for training robotic systems, and Snabbit previously held preliminary talks with Human Archive, but the project did not gain further traction.
Urban Company’s CEO Abhiraj Singh Bhall contradicted such deals on the social network X, which prompted a response from Patel: Urban Company would reportedly be forced to reassess its position to avoid losing relevance due to customer churn. Co-founder Rushil Agarwal noted that Pronto founder Anjali Sardana laughed at him and called him a “fool” when he raised questions about data partnerships; Pronto confirmed the talks but decided not to continue.
In India, other startups are also collecting egocentric data from various work environments, including factory floors. To stand out, Human Archive is developing additional devices: tactile gloves, a full-body motion-capture suit, and wrist-mounted cameras that collect movement data, tactile feedback, and synchronized RGB-D data for sale to AI laboratories. The company believes that video data alone is not enough – combining it with other sensors makes the data significantly more valuable.
“To collect data, we started with the iPhone, then created our own systems and caps. Now we have more than seven different hardware products that we can use interchangeably in various modes. After collecting data from different devices, we are working on synchronizing data from all sources.”
From the perspective of Wing VC partnerships, Zach DeWitt stressed that Human Archive has a unique advantage due to its ability to collect data from multiple sensors. “No one in the world has been able to synchronize and collect RGB-D data from head-mounted displays, force feedback, full-body motion capture, and synchronized chest- and wrist-mounted camera data at scale. They are already conducting internal model training on this data, and leading labs and universities are interested in conducting experiments thanks to the novelty of the sensors and the scale of the dataset they will soon release.”
“The Human Archive network provides immediate, flexible earning opportunities worldwide, lowering the barrier to participation in the AI economy. We see this as a critical bridge between the immediate means of making a living and building infrastructure for a safer, more productive future.”
Despite refusals from some large operators, Human Archive continues to collaborate with smaller startups, offering clients favorable terms. Under the agreement, when a worker arrives at a home, the user can choose a discount in exchange for consent to data collection or pay the full visit fee without recording.
Patel emphasized that most clients opt for the first option, as video footage can help resolve questions about service quality. Workers are offered a baseline rate of $1 per hour for participating in first-person data collection. According to the Economic Times, other companies offer ₹250–₹400 per hour. Patel noted that competitors pay more, but having a company in India allows keeping costs lower.
“The Human Archive network provides flexible and immediate earning opportunities globally, lowering the barrier to participation in the AI economy. We see this as a critical bridge between immediate utilization of labor and building infrastructure for a safer and more productive future,” said Zach DeWitt.
In addition to compensation, there is growing attention to the privacy of data collected via video. The company asserts that its contracts comply with DPDP (Digital Personal Data Protection Act) of India, provides a privacy policy, and explains the purposes of data collection. All data are anonymized, faces are typically blurred. It is also known that India’s Ministry of Electronics and Information Technology is examining consent mechanisms and data-collection practices of startups that work with egocentric data through domestic workers.
Although most data are concentrated in India, Human Archive is gradually expanding its operations in Southeast Asia and the United States. The company is developing a platform that allows anyone to join data collection and get paid. There are also early pilots in the U.S. offering clients cleaning or cooking services in exchange for data from workers.
Overall, several well-funded startups are competing to build physical AI. Such initiatives require large amounts of training data involving people in real-world work – and Human Archive is one of the key players in this field. Whether their approach can scale will depend on partnerships and the uniqueness and volume of data they can collect for physical AI laboratories.
"Proud to share that N1 is backing Human Archive in their $8.2M seed round. For the last years, AI has learned mostly from the internet.Text. Code. Images. Videos. All the digital traces we created online. But the next frontier of AI will need something different. It will need to understand the physical world. How humans cook. Clean. Repair. Build. Carry. Assemble. Move. Improvise. Coordinate. And solve messy real-world tasks. That is the core insight behind Human Archive.
In conclusion, ethical and regulatory aspects remain critical for the development of such initiatives: transparency in data use, workers’ consent, and protecting their rights – these factors will determine the long-term viability of data-collection models in the gig economy and their application in robot training.
