Physician AI Trainer Jobs by Specialty: Pay and Programs

Physician AI trainer jobs by specialty

AI model evaluation jobs for doctors are now posted specialty by specialty. The table lists every program we found recruiting each specialty on 14 November 2026, with pay exactly as the program publishes it. Programs are listed in order of the highest rate shown. A listing that names no rate says so; we do not estimate one.

For what this work is, how it compares with other physician side gigs, and the license and board requirements most listings share, see AI training jobs for physicians. This page covers who is hiring your specialty now.

Mercor shows one hourly figure per listing; Handshake AI, Outlier and DataAnnotation show "up to" ceilings, which most people on a listing do not earn. Listings close often.

Listings open to physicians in any specialty

Some programs recruit doctors as a group for later projects. Open on the same date, alphabetically:

What the work looks like in different specialties

The tasks change with the specialty. micro1's hematology listing describes evaluating "AI-generated transcripts and medical notes from outpatient hematology encounters" (micro1), and its hematology/oncology post asks for comments on "chemotherapy cycles, treatment toxicities, laboratory discussions" (micro1).

Mercor's radiologists "read de-identified patient charts and judge AI answers to clinical questions in radiology" and decide "which of two responses is stronger" (Mercor). Turing's dermatology project asks India-based dermatologists "to review dermatology images" (Turing), and Outlier's surgeons review "model responses and operative videos" (Outlier).

In every case the product is your written judgment on why an answer is unsafe, incomplete or right.

Remote and part-time: hours, location and who qualifies

Hours vary by listing. micro1's hematology and oncology posts say "Part-time, project-based" (micro1). Mercor's pediatrician listing asks for "a minimum of 10 hours/week through mid-November" (Mercor), and OneForma's EHR project lists 10 to 20 hours a week (OneForma).

Location narrows the list. micro1's hematologist post says "Remote, United States only" (micro1), while its emergency medicine post is "open to qualified physicians worldwide" (micro1). Handshake AI requires you to "be based in the U.S." (Handshake AI). Meridial's doctor project is for the United Kingdom and sets the rate "after evaluating your experience, expertise, and geographic location" (Meridial).

If you are looking for non-clinical physician jobs remote because you have stopped practicing, read requirements closely. Mercor's specialty listings ask that you be "Currently practicing and seeing patients in the US" (Mercor), and micro1's hematology/oncology post wants physicians "who have recently practiced outpatient hematology/oncology after fellowship" (micro1). If you are unsure which programs are open to your specialty and country, tell us what you do and we'll email you the programs open to you.

Moonlighting, malpractice and patient data

Residents and fellows: check your program's policy before you start. Training programs use the ACGME definition of moonlighting, "Voluntary, compensated, medically related work performed beyond a resident's or fellow's clinical experience and education hours". At the University at Buffalo, for example, "PGY-1 residents in UB-sponsored, ACGME-accredited programs are not permitted to moonlight", and others must "Obtain the advance written approval of the Program Director" (UB GME). Whether paid AI evaluation counts as medically related work under your program's policy is your program director's call, so ask first.

Attending physicians: read your employment contract for outside-work and conflict-of-interest clauses, and get any required approval in writing. No program page we read mentions malpractice cover. The listings describe evaluation without patient care, but how your policy or employer treats it is a question for your carrier and employer.

Patient data: never upload or paste it. HIPAA says a covered entity "may not use or disclose protected health information, except as permitted or required" (45 CFR 164.502), and information counts as de-identified only under the standard in 45 CFR 164.514. Projects supply their own de-identified material; Mercor says its work "will not involve access to confidential or proprietary information from any employer, client, or institution" (Mercor).

Next step

Compare published rates across all fields in AI trainer pay rates, check a program in are AI training jobs legit?, and read taxes on AI training jobs before your first payment. Nurses and coders have their own guides: AI training jobs for nurses and medical coding AI training jobs.

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).