
Expert pool · Frontier research
Machine learning engineers, data scientists and graduates with a strong ML background. Remote, paid per hour worked, with a short test on real data before paid work starts.
Researchers for RLHF, evaluation and hard reasoning (math, science, multilingual).
Pools group experts by field. When a customer needs people for a task in frontier research, members of this pool are invited first and see the task, its rate and its guidelines before they decide.
Examples of task types in this field. Each task names its exact work before you apply.
How it works
Takes about two minutes. Add your CV and field so we can match you to this pool.
3 to 5 minutes by video and voice. A person reviews it, the AI never rejects you on its own.
Each task in the pool starts with a qualification test on the customer's items. It is unpaid, and so is the calibration round after it.
Once approved, you are paid for measured working time at the task's hourly rate, paid out weekly.
FAQ
Rating and comparing AI answers, grading them against a rubric, writing model answers, trying to make models fail, and labeling images or documents in your field. All remote, in a browser.
Students, graduates and professionals with knowledge in a field that projects need, from law and medicine to engineering and geoscience. You confirm your residence and work authorization during sign-up.
Every role shows its hourly rate before you apply. You are paid for the time you actually spend on paid items, measured by the platform. Rates differ by project and field.
Weekly through Stripe Connect, with Wise as a fallback across the EU and EEA. An instant payout is available for a 1.5 percent fee.
No. The test and the optional practice round are unpaid, and the role page tells you this before you start. The calibration round right after the test, where you get feedback on each item, is unpaid too. All production work after that is paid.
A short live conversation by video and voice, about 3 to 5 minutes, in a language you choose. It is recorded with your consent and stored in the EU. A person reviews the result; the AI never rejects you.
You see which items you got wrong with the reference answer, can practice on your mistakes and try again.
No. You work as an independent contractor under the worker terms and are responsible for your own taxes.
As many as the project has items for you, with no minimum. Availability depends on open projects in your field.
Construction
Construction & building materials
Geosciences
Geosciences & environment
Medicine
Clinical Judgment Annotator
Medicine
Medical QA Specialist
Medicine
Diagnostic Reasoning Expert
Generalist
Generalist tasks
Machine Learning
AI Evaluation Specialist
Law
Legal Reasoning Evaluator
More about paid remote work in this field: guide for frontier research
Email only to start. We invite you as soon as a task in this pool opens.
Apply now