Know a company with years of records? Refer it and earn up to $100K
Guide for experts

AI Training Jobs for PhDs and Postdocs: Pay by Field

Remote part-time work for PhD students, postdocs and PhD scientists: AI training roles in physics, maths, chemistry and bioinformatics, pay and visa rules.

Physics PhD / postdoc
$100-170/hrmicro1, 14 November 2026
Formal maths (Lean 4)
$170-200/hrAlignerr
Typical hours
5-20 a weekHandshake AI listings
Fields in the table
7Physics to bioinformatics
House keys on a lease folder and an inspection checklist
Short answer

The best-paid remote part-time option for most PhD students, postdocs and PhD scientists right now is AI training: writing hard problems in your field, solving them and grading AI answers. On 14 November 2026 micro1 listed physics PhDs and postdocs at $100 to $170 an hour, Alignerr formal mathematicians (Lean 4) at $170 to $200, Handshake AI chemistry PhDs at up to $150, DataAnnotation mathematicians, physicists and chemists at up to $125+, and Mercor physics, chemistry and biology experts at $65 to $105. Most listings ask 5 to 20 hours a week. Students on visas must check their work authorisation first.

  • AI training is the main remote part-time route for PhDs: your research-level judgment is what programs pay for.
  • Published rates run from $35 an hour (subject-matter roles) to $200 (formal proof in Lean); "up to" figures are ceilings.
  • Many listings accept current PhD students; some want a completed PhD or postdoc.
  • F-1 students need CPT or OPT for off-campus work, and your university and funder may have outside-work rules.

Remote part-time options for PhDs: where AI training fits

The traditional side incomes for doctoral researchers are teaching, tutoring, freelance editing and consulting. They still exist, but they mostly pay for teaching or writing skill. AI training pays for research-level judgment: labs need people who can write a problem a strong model gets wrong, prove the right answer and explain where the model's reasoning breaks.

The work is described plainly on the listings. micro1's physics role asks contributors to "Produce and validate high-quality solutions to advanced theoretical and mathematical physics problems", and may assign you as a solver, auditor or adjudicator (micro1 listing). Outlier says it has onboarded "700K+ PhD, MS students" (Outlier).

AI training jobs for PhDs by field

Every program listed had a live listing for the field on 14 November 2026. Rates are per hour as published; "up to" figures are ceilings.

FieldPrograms listing it now (alphabetical)Published payWho they ask for
PhysicsAlignerr; DataAnnotation; Handshake AI; Mercor; micro1; xAI$35-55; up to $125+; up to $75; $70-100; $100-170; $45-75 (US candidates)Advanced degree; micro1 asks for a PhD or postdoc; Handshake AI PhDs and postdocs
Mathematics and statisticsDataAnnotation; Handshake AI (math and statistician roles); Invisible (India); Mercor; micro1; xAIUp to $125+; up to $75; $8-65; $40-75 and $90-140 (US, PhD); $60-120; $45-75Master's or PhD on most; Mercor's domain role requires a PhD
Formal mathematics (Lean, Coq, Isabelle)Alignerr; Mercor$170-200; $80-130 (US)Master's or PhD with proof-assistant experience
ChemistryAlignerr; DataAnnotation; Handshake AI; Mercor; micro1$35-55; up to $125+; up to $150; $65-105; $45-90Handshake AI PhDs and postdocs; micro1 PhD
Computational biology and bioinformaticsDataAnnotation; Handshake AI; Mercor (genomics); micro1Up to $125+; up to $75; $70-100; $45-90DataAnnotation accepts current PhD candidates; Mercor encourages recent PhDs and postdocs
BiologyHandshake AI; MercorUp to $125; $70-105Handshake AI master's, PhDs and postdocs
Any STEM fieldAlignerr (Masters or PhD Researcher); Handshake AI (STEM PhD, EU and UK)$75-90; up to $48Master's or PhD; EU or UK residence for the Handshake role

Sources, read 14 November 2026: Alignerr physics, chemistry, mathematician, Lean 4 and PhD researcher listings; DataAnnotation mathematician, physicist, chemist and computational biologist roles; Handshake AI physics, math, statistician, chemistry, biology, bioinformatics and STEM PhD (EU) listings; Invisible mathematics listing; Mercor physics, mathematics, mathematics domain, formal mathematics, chemistry, biology and genomics listings; micro1 physics, mathematics, maths and statistics, chemistry and computational biology listings; xAI math and physics listings.

Snorkel's math community also looks for "Master's or PhD-level expertise" and is open only to eligible US-based contractors outside California (Snorkel).

Math AI trainer and physics AI tutor roles

Math AI trainer and physics AI tutor roles usually mean the same work: teaching a model, not students. xAI's AI Tutor roles in math and physics pay US-based candidates $45 to $75 an hour, are open remotely in the United States, United Kingdom, Canada, Australia and India, and prefer full-time availability (xAI math listing).

The top of the range is formal proof. Alignerr's mathematician roles pay $170-200 an hour to "Formalize advanced mathematical arguments and theorems within Lean 4" and ask for a master's or PhD in mathematics (Alignerr); Mercor's formal mathematics role pays $80-130 an hour in the US for Lean, Coq or Isabelle work (Mercor). If you want tutoring of real students instead, see AI tutor jobs.

How many hours, and does it fit a PhD?

Most listings are built around a research schedule. Handshake AI's physics, math, chemistry and biology listings say most contributors work approximately 5-20 hours per week on an active project, with no minimum (Handshake AI); its EU and UK STEM PhD role mentions about 10 to 20 hours (Handshake AI). Alignerr lists 10-40 hours/week (Alignerr). Mercor's genomics and formal mathematics roles ask for at least 20 hours a week on weekdays (Mercor), which is a lot next to a thesis.

Work comes in projects, so expect busy weeks and empty ones. Do not plan rent around it.

Visas, university rules and funding: check before you sign

International students in the US: USCIS says "F-1 students may not work off-campus during the first academic year", and that after it they may work off-campus through Curricular Practical Training or Optional Practical Training, authorised in advance (USCIS). Handshake AI says F-1 students "who are eligible for CPT or OPT may be eligible for projects", and that "STEM OPT is not supported" (Handshake AI). Alignerr says visa holders can take part only if their visa permits contract work, and it does not sponsor visas (Alignerr FAQs). Your designated school official, not a program, decides what you may do.

Outside the US, student and researcher visas set their own limits on hours and self-employment. Check yours with your university's international office.

Universities and funders often have outside-work and conflict-of-interest policies for doctoral students and postdocs, and some fellowships limit paid work. Ask your supervisor or graduate school before you commit hours, and keep unpublished results, lab data and anything under your group's agreements out of every task.

Next step

For field pages, see AI training jobs for software engineers if you are in CS or ML, and AI evaluation jobs for what grading work looks like. Pay is contractor income: see AI training jobs and taxes.

Tell us what you do. We'll send you the paid AI work that fits your field, wherever you live. It's free.

Origin Data Partners is independent of the programs on this page. A program may pay Origin when Origin introduces someone who joins it, which never changes what you are paid (how Origin is paid).

Common questions

Can PhD students do AI training work?

Yes, many listings accept current doctoral students. Handshake AI's physics, math and chemistry roles name PhDs and postdocs, and DataAnnotation's computational biology role accepts current PhD candidates. Check your visa, university policy and funding terms first.

Can international PhD students on an F-1 visa do AI training work?

Only with work authorisation. USCIS says F-1 students may not work off-campus in their first academic year and need CPT or OPT after that. Handshake AI says CPT- or OPT-eligible F-1 students may be eligible but STEM OPT is not supported. Ask your designated school official.

Which fields pay PhDs the most for AI training?

On 14 November 2026 the highest published rates were Alignerr's formal mathematics roles at $170 to $200 an hour and micro1's physics PhD and postdoc role at $100 to $170. Handshake AI listed chemistry PhDs at up to $150.

What is a math AI trainer?

Someone who writes hard maths problems, solves them and grades how an AI model solves them, step by step. Listings at DataAnnotation, Handshake AI, Mercor, micro1 and xAI asked for a master's or PhD on most roles.

How many hours a week do AI training projects for PhDs take?

Handshake AI says most contributors work about 5 to 20 hours a week on an active project, with no minimum. Alignerr lists 10 to 40 hours, and some Mercor roles ask for at least 20 hours a week on weekdays.

Do postdocs need permission for AI training side income?

Often, yes. Postdoctoral appointments and fellowships commonly have outside-work and conflict-of-interest rules, and visas may limit work. Check your appointment letter and ask your institution before you start.

Sources and further reading

For experts

Tell us what you do

Origin works with AI training programs that hire nurses, doctors, lawyers, accountants, engineers, teachers and other professionals as contractors. Tell us what you do once, and we’ll email you the paid AI work that fits.

  • Free for experts: you never pay Origin anything.
  • Nothing goes to a program unless you agree to an introduction.
  • Each program decides who it accepts and what it pays. Acceptance and work are not guaranteed.

Compare what each program publishes about pay

Free for experts

Get paid for your expertise

Takes two minutes. Tell us what you do and we’ll send you paid AI work that fits.

Add more detail · Optional, helps Origin reply
Kinds of work you want

What is your company’s data worth?

AI companies pay $20K to $5M for company records. Six questions, no files.