Overview
Frontier models still fail at rigorous multi-step mathematics: invalid derivations, unjustified leaps, and confidently wrong final answers. Automated checks can't grade a proof. You can.
As a Mathematician you'll evaluate and correct mathematical reasoning across everything from competition problems to graduate-level material, and your judgment becomes the standard the models learn from.
What you’ll actually do
- Craft problems that expose weak mathematical reasoning, from olympiad-style puzzles to graduate-level proofs.
- Review AI solutions step by step: catch invalid derivations, flag unjustified leaps, and rate the quality of the reasoning.
- Write the gold-standard solution when the model falls short, with the rigor of a referee report.
Roles this fits
Common backgrounds: Mathematician, Applied Mathematician.
What we look for
- Strong background in pure or applied mathematics through graduate study, competition math, or equivalent experience.
- Fluency writing precise, well-structured proofs (LaTeX comfort helps).
- Clear written English: your explanations are the training signal.
- No degree required. We care about what you can do, not where you learned it.