Chinese experts turn to freelance gigs to train AI models

Chinese architects, engineers, and lawyers facing a slow economy and reduced pay are taking piecework data roles to train artificial intelligence models in their own fields.
Major technology corporations in China are directly signing up experienced specialists to build advanced data sets.
Faced with layoffs and pay cuts, these skilled workers treat complex data annotation assignments as an immediate way to secure supplementary cash.
The practice mirrors American operations such as Mercor and Handshake that hire technical specialists to narrow the gap with Silicon Valley.
Cuicui, a Shenzhen architect in her 40s who spent decades designing public schools, hospitals, and stadiums, saw her compensation plunge by 50% over five years as state infrastructure allocations dried up.
She registered on TalentsAI to fund her mortgage along with her child’s after-school activities, spending her evenings feeding instructions to an algorithm on commercial feasibility and architectural designs.
“It is just like teaching a child,” Cuicui said, explaining that every instruction she writes comes directly from her professional assignments.
Entities in the United States including Mercor, Surge AI, and Handshake previously contracted thousands of scientists, film directors, and legal counselors to refine corporate systems.
China’s National Data Administration intervened in June by calling for higher knowledge density in digital sets, formally pushing college graduates and veteran industry workers into data annotation.
Specialized contractors confirmed that while local compensation remains lower than American rates, taking these demanding tasks provides vital income during a severe white-collar job crunch.
New platforms hire professionals for corporate systems

Siriser, launched by Alibaba in January, contracts mechanical engineers, music composers, and teachers, while ByteDance’s Xpert platform gathered 50,000 specialists by 2025 alongside independent portals like TalentsAI and MeetChances.
MeetChances published a direct recruitment message in June on the social platform RedNote, asking what downsized middle-aged professionals could do besides joining the algorithmic workforce.
Specialized data has become the main battleground since Chinese laboratories closed the gap with Silicon Valley in core coding and logical reasoning, according to Sydney-based artificial intelligence researcher Tiezhen Wang.
ByteDance rolled out Doubao Work and Tencent launched WorkBuddy to generate presentation files and digital pages. “Making models perform better at work will be a very strong differentiator in the future AI competition,” Wang said.
Freelancers craft real-world occupational tests and supply actual workflow records without using automated software themselves, earning between 100 and 500 yuan ($15 to $74) for assignments that require several hours of concentration.
Work is getting harder.
Corporate guidelines at Xpert outline that unapproved submissions yield zero payment, while Siriser maintains a formal review and dispute procedure for denied payouts.
Economic indicators show youth unemployment reached 17.9% in July, pushing white-collar workers toward contract labor as massive graduating classes flood the market.
Liu, a Shenzhen legal professional holding a master’s degree, bypassed conventional law firms to avoid intense hours, low entry pay, and workplace harassment.
Working full-time at a legal tech company, Liu takes evening assignments from Moonshot AI, Tencent, and TalentsAI to assess how software formulates judicial decisions and client advice.
“If I don’t pass my knowledge to AI, others will do that,” Liu stated, adding that she is studying for the civil service exam to find greater employment security.
Data collection expands from basic tasks to enterprise needs
For ten years, mainland annotation enterprises relied on low-wage groups like rural workers, stay-at-home mothers, and student interns to mark inputs for acoustic tools and autonomous cars.
Research group IDC reported in August that enterprise software needs are overtaking consumer tasks, projecting the domestic training data sector will climb 25% from 2025 to reach 7.8 billion yuan ($1.1 billion) in 2026.
University of Toronto associate professor Julie Yujie Chen noted that harsh employment conditions drive middle-class specialists toward data labor after official agencies legitimized the field.
“The middle class in China with ordinary white-collar jobs are living in a state where they face this urgency to find a supplementary income,” Chen said.
Jessica, a veteran Shanghai software engineer with twenty years of practice, entered data annotation in July after Claude called the work an immediate entry point into algorithmic technology.
Her contracts involved evaluating extended text processing and testing autonomous software capabilities across specialized scenarios.
The small income barely offset the hours invested or the risk of exposing project files, but corporate staff cuts and contractor replacements convinced her to maintain side options.
“You never know how long you will get to keep your job,” she said.






