He was hired by three companies based in Argentina, India and the United States, each paying between US$3 and US$10 per hour to film chores such as dishwashing, cleaning, cooking, making the bed, ironing, folding clothes and sweeping.

Egocentric training data are sensor or video recordings made from the perspective of the person doing the task, usually captured with a head‑ or chest‑mounted camera that keeps the worker’s hands in view.

Sergio Bruccoleri, vice‑president of delivery for Appen and based in Barcelona, said the Australian ASX‑listed firm can collect about 60,000 hours of egocentric video each month from up to 5,000 contributors worldwide, with China currently the biggest market for robotics data.

Bruccoleri added that Appen is now focusing more on higher‑value egocentric data from wealthier countries such as the United States and Canada, because the first robots will be sold there and appliances – like dishwashers – differ between markets.

Many contributors, including Dunggio, report problems such as video rejections, a US$1 transfer fee for withdrawals under US$10, and concerns about filming inside private homes. Despite these issues, most see the work as a flexible side‑hustle.

These first‑person videos give robots a clear view of human hands and movements, allowing AI systems to learn how to manipulate everyday objects. The more varied the recordings, the better robots can adapt to different homes in the future.