For mathematicians

Every major frontier lab is now investing heavily in mathematical reasoning. The reason is not that mathematics is commercially urgent by itself. The reason is that mathematics is the cleanest test for whether a model can think through a problem carefully, catch its own errors, and produce a chain of reasoning that survives inspection.

That inspection is where trained mathematicians come in. A model that produces a proof-shaped answer is not the same as a model that produces a correct proof. The difference is invisible to almost everyone except someone who has spent years doing this. Frontier labs know this, and they have quietly built expert-annotation programs for mathematics that pay accordingly.

What the work looks like

• Proof verification. You are given a mathematical claim and a model's attempted proof. You verify it, correct it, or explain why it fails. The failures are often subtle.
• Problem authoring. You write problems at a specific level (olympiad, undergraduate, graduate) with clean solutions, dense enough for training use.
• Adversarial testing. You construct problems that look easy but expose a specific reasoning failure mode. This is where research training compounds fastest.
• Rubric design. You author the criteria a mathematician would use to grade a proof, and audit the criteria the lab is currently using.

The work is asynchronous. Sessions are typically two to four hours. Structured deliverables.

What labs actually look for

Graduate training in mathematics, statistics, theoretical computer science, or a closely related field. Active engagement with the field, either through research, teaching, or serious independent work. Written English at the level a research paper expects.

They do not look for machine learning experience. Very few mathematicians hired for this work have ever trained a model. They are hired for the depth of their reasoning, not their familiarity with the tooling.

Where the pay lands

Rates are in USD and are set well above what local academic salaries typically pay in most countries. For mathematicians in emerging markets, this often means a single afternoon of expert work earns what a week of adjunct teaching does. The gap is real and it is deliberate.

How to enter

Sign up, upload a recent CV that shows your training and current activity, and let the platform surface open roles in mathematical annotation, evaluation, and problem authoring. Apply where you can deliver rigorously. The frontier is moving; it needs mathematicians who care about the details.