Work onreasoning.
Create mathematical tasks, investigate reasoning errors and build data that AI teams can train and evaluate with.
We welcome expressions of interest for future project work. Scope, timing and compensation are agreed for each engagement.
Areas of work
Where you
could contribute.
01Mathematical Problem Designer
Design original problems, develop reasoning task families and write complete reference solutions. Help define what makes a problem a useful test of a particular skill.
Relevant experience
- Olympiad, graduate-level or research mathematics.
- Original problem creation and clear mathematical writing.
- Explaining assumptions and checking alternative solutions.
02Mathematical Reviewer
Independently work through tasks and solutions. Identify invalid steps, ambiguous wording and gaps in grading criteria, then provide specific feedback.
Relevant experience
- Rigorous mathematical problem-solving and proof review.
- Finding counterexamples and checking edge cases.
- Writing feedback another expert can act on.
03Lean 4 Formalization Specialist
Translate mathematical statements and proofs into Lean 4. Check that the formal statement captures the intended claim, and document dependencies and proof-checking results.
Relevant experience
- Lean 4 and its mathematical libraries.
- Readable formal statements and kernel-checked proofs.
- Understanding axioms, incomplete proofs and library versions.
04Research Engineer — Synthetic Data
Build generation and evaluation tools that turn a task specification into a reviewable dataset. Work on task diversity, dataset lineage and reproducible model evaluations.
Relevant experience
- Python and practical LLM evaluation.
- Generation pipelines, data checks and reproducible experiments.
- Measuring accepted output, failure modes and review cost.
05Expert Project Lead
Coordinate problem writers and reviewers. Translate a research brief into clear acceptance criteria, keep review decisions consistent and track delivery quality.
Relevant experience
- Leading expert data or mathematical content projects.
- Quality review, feedback and disagreement resolution.
- Clear communication with technical clients and contributors.
These profiles describe potential project collaborations. Introductions help us understand your expertise; specific work and terms are discussed when a matching project is available.
A different
kind of expertise?
Tell us what you know and show us work you are proud of.