What the work involves
Maths AI training means writing hard problems and checking a model's working, step by step. Mercor's maths listing describes writing a problem at research or competition level with its intended solution, grading model-written proofs on whether each step follows, and flagging where a model reaches a correct answer through an argument that does not prove it.
Where math experts can apply
- Mercor, Mathematics Domain Expert (US). A specific opening, open when we checked on 8 November 2026: $90 to $140 an hour for US-based mathematicians with a PhD in mathematics, applied maths, statistics, operations research or a close field. It is a senior role: it asks for a faculty appointment, principal investigator, staff or senior scientist post (or equivalent senior research role) and a strong peer-reviewed publication record. See the listing.
- Mercor, mathematics pool. A standing listing, which Mercor says is "not a specific job opening". Most past contracts in this field paid $40 to $75 an hour. Asks for a graduate degree in maths or a related field, or an equivalent research or competition background. See the pool.
- DataAnnotation, STEM tier. Starting at $40 to $125+ an hour for work that needs a master's or PhD in maths or a science, or a bachelor's plus 10+ years of experience. Read the FAQ.
- Outlier. Lists "Math Expert" among its roles, with undergraduate level expertise as the minimum and graduate study preferred. It does not publish a rate. See Outlier.
- Handshake AI. US listings for maths PhDs paid $75 an hour and $85 to $110 an hour. Both were closed when we checked.
What you need to apply
Most paid maths work we found asks for graduate study. Mercor's pool accepts a graduate degree or an equivalent research or competition background, its US role needs a PhD plus a senior research post and publications, and DataAnnotation's STEM tier needs a master's or PhD (or a bachelor's with 10+ years' experience). Outlier sets a lower minimum of undergraduate level expertise.
Mercor's listings say it cannot support US H-1B or STEM OPT candidates.
What it pays
Published ranges run from $40 an hour (the bottom of past Mercor pool contracts and DataAnnotation's starting point) to $140 an hour (top of Mercor's US PhD role). PYMNTS reported in October 2026 that a Princeton PhD placed by Mercor worked a $70 an hour role.
Things to know before you start
- This is independent contractor work, with no benefits or guaranteed hours.
- Competition maths listings fill and close fast. Several Mercor competition and olympiad listings were closed when we checked.
- Expect an assessment before any paid work, and check whether it is paid. One Outlier reviewer reports losing hours of unpaid work (Trustpilot, 4 November 2026).
More by profession
- AI training jobs for coders
- AI training jobs for nurses and doctors
- AI training jobs for lawyers
- Best data annotation jobs, all platforms compared
Questions
Do I need a PhD?
Not always. Mercor's US domain expert role needs one, plus a senior research post, but its general maths pool accepts a graduate degree or an equivalent research or competition background. DataAnnotation's STEM tier accepts a master's, and Outlier's minimum is undergraduate level expertise.
How much do math experts earn training AI?
Published ranges we found run from $40 to $140 an hour. The highest figure on an open listing on 8 November 2026 was Mercor's senior US role for maths PhDs at $90 to $140 an hour.
Sources
- Mercor listing: Mathematics Domain Expert (US)
- Mercor listing: Mathematics Expert
- DataAnnotation FAQ (pay tiers and requirements)
- Outlier home page
- Handshake AI: Mathematics PhDs, AI Trainer (closed; Base10 job board copy)
- Handshake AI: Math Expert (closed; Base10 job board copy)
- PYMNTS: AI labs pay lawyers $400 an hour to train models (20 Oct 2026)