This startup is betting that India’s economy can train the world’s robots

Over the past few years, India’s online food delivery market has grown exponentially, with both Zomato and Swiggy going public and the number of cloud-based kitchens growing. Meanwhile, startups working on home services, such as on-demand staffing platforms like Urban Company, Snabbit, and Pronto, have gained popularity.

Silicon Valley-based startup Human Archive is getting in on the trend, partnering with these companies to have workers wear special caps with cameras to collect egocentric (first-person) video data of daily activities that can be used to train robots.

Without naming specific partners, the startup said it is working with companies in the home services, hotel, and restaurant sectors to collect egocentric data, and says it has more than 1,000 active headsets spread across multiple locations.

After that, Human Archive said on Tuesday that it raised $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.

The startup was founded by three students from UC Berkeley and one from Stanford – Samay Maini, Rushil Agarwal, Shloke Patel, and Raj Patel, the two are cousins. (Raj Patel is the CEO.) All four have research backgrounds that include robotics, hardware, and tactile data.

The company’s innovation is a direct bet on where the AI ​​industry is headed. As robotics labs and frontier AI companies rush to build machines that can perform physical tasks in the real world, they are facing a critical bottleneck – a lack of high-quality, real-world training data that shows people doing everyday work. The Human Archive’s bet is that workers in India’s developing economy represent an untapped and untapped source of such data.

While Human Archive is working with many partners, the startup said it has been rejected by several local Indian companies, including Pronto and Urban Company, for collaboration.

The company’s rejection by major players became public fodder last weekend, when Indian company Entrackr reported that Pronto wanted to partner to collect employee data for robot training and that Snabbit had early talks with Human Archive before the project fell apart.

Urban Company CEO Abhiraj Singh Bhal responded with X, saying that the company would not engage in such schemes – prompting Patel to respond that Urban Company would soon be forced to rethink or risk losing relevance to customers. Founder Rushil Agarwal was unmoved, writing that Pronto co-founder Anjali Sardana laughed at him and called him “stupid” when he brought up the idea of ​​data collaboration. Pronto agreed to the talks but chose not to go ahead.

Across the country, some startups are collecting egocentric data from different workplaces, including factory floors. To differentiate itself, the Human Archive uses and develops additional devices, such as tactile gloves, a full-body imaging suit, and wrist cameras to capture data, including movement and touch force, aligned in RGB-D (real-time color image matching and depth information), for sale to AI labs. The startup believes that video data alone is not enough but combining it with other sensor data makes it more valuable.

Initially, the Human Archive used temporary setups or off-the-shelf rigs to capture data. Now it runs on custom hardware that works together and captures different types of data. It already has more than 50 different devices used to collect different data points.

“To capture data, we started with iPhones; then we built our own custom rigs and caps. Now we have over seven different hardware products that we use interchangeably in all the different ways. After the data was collected from the different devices, we worked to synchronize the data from all these different sources,” Patel said on the phone.

The company said it is developing ways to fine-tune AI models with its own data and test them on robots to test job performance. By doing this, startups can demonstrate the quality of their data to potential customers and train internal models.

Zach DeWitt, a partner at Wing VC, said the startup has the unique advantage of collecting data from multiple sensors.

“No one else in the world has been able to synchronize and collect RGB-D headsets, respond to feedback, capture full body movements, and chest and hand camera data at scale. They have been doing internal model training with this data, and every major lab and university is interested in conducting experiments on it because of the novelty of the sensors and the scale of the new data set,” he is told again.

Data collection in India and expansion plans

Despite being rejected by major players in the home services industry, Human Archive has partnered with small startups to offer discounted services to customers. When an employee comes to the home, consumers are given a choice through the app: pay a discounted price if you agree to data collection, or pay full price for an unrecorded visit.

Patel said customers are happy with the first choice, as disputes about service quality are common, and video recording can help resolve them.

The company pays employees a base rate of $1 an hour for participating in self-reported data collection. A report from the Economic Times suggests that some companies pay R250 to R400 an hour (about $2.63 to $4.20). Patel said competitors pay more than Human Archive, but its presence on the ground in India allows it to keep compensation low.

“The Human Archive network provides fast, flexible monetization opportunities around the world, lowering the barrier to participation in the AI ​​economy. We see this as an important bridge that supports a fast-paced life while building the infrastructure for a safe and productive future,” said DeWitt.

Besides the payment of wages, there are privacy concerns about data collection through video recording. It is unclear what information the Human Archive provides about how their images are used. The company said its commercial contracts comply with the Indian Data Protection Act (DPDP), as it displays a privacy policy notice, and consent details explaining the purpose of data collection and how it is processed. The company said all data is anonymized and faces are blurred due to recording. Last week, Moneycontrol reported that India’s Ministry of Electronics and Information Technology is looking into consent and data collection methods for startups that collect egocentric data through domestic workers.

Although Human Archive mainly collects data in India, it has started expanding to Southeast Asia and the US. The company is also creating a platform for anyone to participate in data collection and earn money. It also wants to offer customers in the US services such as cleaning or cooking to receive data collection by participating employees – although these programs are in the early stages of testing.

Many well-funded startups are racing to build real-world AI. Doing so requires a large amount of training data that shows people at work – and Human Archive is one of the players competing to fulfill that need. Whether its approach can scale will depend on the partnerships it strikes and the volume of data it can collect to satisfy the appetite of portable AI labs.