Post-00s woman from Zhejiang, Yang Xinyi, rents 30 apartments in Wuhan, leads a team of over 50 people teaching humanoid robots to do housework

To teach robots to do housework, 22-year-old Yang Xinyi did something that left real estate agents baffled: she asked to rent 30 apartments in one go, living in each for only about a week. "Why are you renting so many apartments?" the agents pressed. She explained that the apartments weren't for people to live in, but for robots to "practice" in. Many agents couldn't make sense of this reason. Some refused outright, some suspected she was a scammer, and others demanded she provide proof of employment, a work badge, and her ID card. Recalling the experience in June 2026, Yang Xinyi was caught between laughter and tears. This young woman from Zhejiang, born in 2004, is now a project leader at a tech company in Wuhan. She leads a team of more than 50 people collecting training data for humanoid robots. From teaching robots to pick up and put down an object to trying to get them to tidy rooms, make coffee, and wash laboratory glassware, she and her colleagues do something every day that seems simple but is actually complex: teaching robots, bit by bit, the life skills humans take for granted. Starting by teaching robots to pick things up, and also judging whether they do it right In October 2025, Yang Xinyi moved from Zhejiang to Wuhan, Hubei, to join the operations team of a tech company's humanoid robot innovation center. With a PICO headset on and two controllers in hand, the robot in front of her can be guided to complete specified actions. Yang Xinyi's job is to repeat these operations, turning the motion process into training data for the robots to learn from. At first, the training content was just picking up and putting down objects. As the tasks progressed, the training scenarios also expanded from simple workbenches to complex environments such as supermarkets, homes, pharmacies, kitchens, and restaurants. In supermarket scenarios, robots need to learn operations such as stocking shelves, restocking, putting items on shelves, and packing; in home scenarios, they try to store items and tidy rooms; in the kitchen, pouring water, fetching items, and making coffee become new tasks. "Whatever ordinary people do, we teach robots to do," Yang Xinyi explained. Only after really getting hands-on did she realize that a casual human action can be a difficult problem for a robot. For example, a person can hold a phone in the left hand and pick up a cup with the right, completing both actions at once. But robots cannot yet handle such two-handed coordinated operations as flexibly as humans. Trainers must break down continuous actions and collect data one by one. Not only must they teach robots to perform actions, they must also judge whether they do them correctly. After becoming a reviewer, Yang Xinyi began checking whether training data was qualified and whether the robot's completion of actions met requirements. As the project advanced, she successively served as team leader and project leader, and now leads a team of more than 50 people working full-time as teachers for robots. Renting 30 apartments for robots: real estate agents thought they had met a scammer As training tasks increased, Yang Xinyi and her team encountered a new problem: robots needed to learn more and more skills, and the training environments had to be more varied. She explained that in addition to operating the robot body itself, they can also use bodyless data collection devices, simulating grasping motions through a headset and handheld grippers; they can also use devices with only a robotic arm to specifically train hand operations. Different collection methods each have their advantages, but for robots to adapt to complex real environments, repeating actions in a fixed venue is not enough. In June this year, the team received a data collection order that required entering different real home environments to collect the data robots need for "doing housework." This time, what stumped Yang Xinyi was not the robots but the apartments. Because the task required different layouts, different spatial scales, and different furniture arrangements, the team planned to find about 30 homes with signs of actual habitation, using each for about a week. During the day, she had to handle project work at the company; whenever she had time, she contacted real estate agents to find listings. But when they heard the lease was only for a week, many agents lost interest. Later, Yang Xinyi simply pulled the agents she had contacted into a group chat and explained the need uniformly: they wanted to rent 30 apartments for robots, the lease term was short, and the places had to show signs of people having lived there. Unexpectedly, the more detailed her explanation, the more people doubted her. "They said they had never heard of renting apartments for robots," Yang Xinyi recalled. Some asked her to show proof from her employer, and others questioned that she was too young, thinking such a big project could not possibly be handled by a young woman. Just when the listings seemed hopeless, a woman in her forties lent a hand. This agent, an older sister figure, was very interested in robots; she not only helped her contact more than 10 apartments but also came to the site to watch the team conduct data collection. "She thought it was very interesting and was very willing to help us." Speaking of this experience, Yang Xinyi is still grateful. When will robots enter the home? She is also waiting for the answer During a year of work, Yang Xinyi has heard many questions. Some people ask directly: "When will robots be able to enter ordinary homes and cook and tidy rooms for people?" In her view, current robots can already complete tasks in some specific scenarios, but there is still a long way to go before they can adapt to the complex and ever-changing living environments of thousands of households. The same object placed in a different position; the same room with a changed furniture layout—both can pose new challenges for robots. That is precisely why a single task requires repeated data collection, and those seemingly boring data gain meaning. She believes that as data accumulates and technology iterates, the tasks robots can complete will gradually increase. Her family is also very curious about her profession. Her younger brother once asked curiously, "What if the Earth is taken over by robots in the future?" Yang Xinyi laughed and used videos taken at work to explain to him that real robots are not as omnipotent as in science fiction movies; behind every action they complete is a large amount of research and development and training. Now, her parents have a better understanding of her work and are proud that their daughter can participate in this emerging industry. "I feel that I entered this industry relatively early, and being able to participate in and witness its development is a very meaningful thing!" Yang Xinyi said she is willing to keep doing this work, because she can see changes in robots every day.