This Startup Seeks to Harness India’s Gig Economy for Global Robot Training

In recent years, the online food delivery industry in India has experienced significant growth, with Zomato and Swiggy going public and cloud kitchens gaining traction. Additionally, startups focused on home services, like Urban Company, Snabbit, and Pronto, have surged in popularity.

Silicon Valley-based startup Human Archive is leveraging this trend by partnering with these companies to provide workers with specialized caps equipped with cameras, allowing the collection of egocentric (first-person perspective) video data from daily tasks, which can be used to train robots.

While not naming specific partners, the startup noted that it is working with businesses across the home services, hostel, and restaurant sectors to collect egocentric data, claiming to have deployed over 1,000 active headsets in various sites.

Building on this momentum, Human Archive announced on Tuesday that it has raised $8.2 million in funding from Wing Venture Capital, NVP Capital, Y Combinator, and angel investors from OpenAI, Nvidia, Google, Mercor, AfterQuery, BAIR, SAIL, Brad Boa, and Meta.

Founded by two Berkeley and two Stanford students—Samay Mani, Rushil Agarwal, Shloke Patel, and Raj Patel, who serves as CEO and is cousins with Shloke—the team brings research expertise in robotics, hardware, and tactile data.

The company’s foundation is strategically aligned with the future direction of the AI industry. As robotics labs and advanced AI firms seek to create machines capable of performing physical tasks in real-world scenarios, they face a critical challenge: a shortage of high-quality, real-world training data depicting humans engaged in everyday activities. Human Archive sees the workforce behind India’s booming gig economy as an untapped and scalable resource for this data.

Although Human Archive is collaborating with various partners, it has faced rejections from several prominent Indian home services firms, including Pronto and Urban Company, for potential collaborations.

This public resistance emerged recently when the Indian outlet Entrackr reported that Pronto is actively seeking partnerships to collect worker data for robotics training and that initial discussions with Snabbit had taken place before ultimately not materializing.

Urban Company CEO Abhiraj Singh Bhal responded on X, stating that the company would not pursue such partnerships. In response, Patel suggested that Urban Company might need to reconsider to remain relevant amid customer turnover. Co-founder Rushil Agarwal recounted that Pronto’s founder Anjali Sardana had ridiculed him and labeled him “stupid” when he proposed a data partnership. Although Pronto confirmed discussions, they chose not to continue.

Other startups nationwide are also gathering egocentric data from various workplaces, including factories. To differentiate itself, Human Archive is developing and using additional devices, such as tactile gloves, a full-body motion capture suit, and wrist cameras to capture data on motion and tactile force, all synchronized with RGB-D (real-time color imagery combined with depth information), aimed at selling to AI laboratories. The startup believes that relying solely on video data is insufficient and that supplementing it with additional sensor data significantly enhances its value.

Initially, Human Archive utilized makeshift setups and commercial rigs for data collection. It is now advancing towards custom hardware that integrates multiple data types. The company operates over 50 diverse devices to collect various data points.

“To capture data, we began with iPhones and progressed to custom rigs and caps. Currently, we have more than seven different hardware products that we utilize interchangeably across various modalities. After gathering data from these different devices, we focused on synchronizing the information from all these diverse sources,” Patel explained during a call.

The company is also working on methodologies to fine-tune AI models using its data and testing them on robots to evaluate task efficiency. This approach enables the startup to demonstrate the quality of its data to potential clients while training internal models.

Zach DeWitt, a partner at Wing VC, emphasized the startup’s unique advantage in collecting multi-sensor data.

“No other organization globally has succeeded in synchronizing and collecting headset RGB-D, force feedback, full-body motion capture, and chest and wrist camera data at scale. They’ve been conducting internal model training on this data, and every major lab and university is eager to conduct experiments due to the uniqueness of the sensors and the forthcoming release of this new dataset,” he informed TechCrunch.

Data Collection in India and Expansion Aspirations

Despite experiencing rejections from key players in the home services sector, Human Archive has partnered with smaller startups to offer customers discounted services. When a worker arrives at a home, consumers can choose via the app to either pay a discounted rate in exchange for consenting to data collection or pay the full price for an unrecorded visit.

Patel observed that customers typically prefer the former option, as disputes over service quality are frequent, and video recordings can effectively resolve such conflicts.

The company compensates workers with a base rate of $1 per hour to participate in egocentric data collection. According to a report from the Economic Times, other companies offer between ₹250 and ₹400 per hour (approximately $2.63 to $4.20). Patel acknowledged that while competitors pay higher wages than Human Archive, its local presence in India allows it to maintain lower compensation rates.

“Human Archive’s network creates immediate, flexible earning opportunities worldwide, reducing barriers to participation in the AI economy. We see this as a critical link that supports immediate livelihoods while laying the groundwork for a safer, more productive future,” DeWitt stated.

Beyond wage considerations, there are privacy concerns surrounding video recording data collection. The details of how Human Archive informs workers about the use of footage remain unclear. The company asserts that its commercial agreements comply with India’s Digital Personal Data Protection (DPDP) Act, displaying a privacy policy notice along with consent information that explains the purpose of data collection and processing. Human Archive claims all data is anonymized, with faces blurred in recordings. Recently, Moneycontrol reported that India’s Ministry of Electronics and Information Technology is reviewing the consent mechanisms and data collection practices of startups collecting egocentric data through home service workers.

While Human Archive primarily gathers data in India, it has commenced expansion into Southeast Asia and the U.S. The company is also developing a platform for anyone to engage in data collection and earn money. Additionally, it aims to offer services such as cleaning or cooking in the U.S. in exchange for data collection by participating workers, although these initiatives are still in early pilot stages.

Numerous well-funded startups are vying to develop physical AI, a task that requires vast amounts of training data illustrating humans at work, and Human Archive is one of the contenders striving to meet that demand. The scalability of its approach will rely on the partnerships it fosters and the uniqueness and volume of data it can amass to satisfy the requirements of physical AI labs.