Radiologist (MD): Medical Imaging AI Annotation & Evaluation

What you'll do

This is non-clinical work for radiologists: no patients and no live diagnosis. Instead you produce the ground truth that imaging AI systems are trained and graded against. Mercor runs it as a shared expert pool, so after onboarding you can be matched to any of several concurrent imaging workstreams by subspecialty, availability and interest, and move between them as priorities change.

The ad lists six kinds of work you may be given:

• Case interpretation with reasoning: working through cases that pair studies with clinical context, writing out your reasoning step by step, not just the impression
• Structured labelling: marking discrete findings against a schema (presence, location, laterality, severity, interval change)
• Reference reports: writing the report a study should produce
• Output review: grading AI-generated reads, flagging hallucinated or missed findings and unwarranted certainty
• Rubric and guideline writing: defining what separates an excellent read from an acceptable one
• Hard-case writing: building imaging questions designed to find the limits of current models

Tasks run from about 10 minutes for structured labelling to about an hour for full case authoring, and each stream sets its own throughput target. Being able to move between streams is useful: if one project winds down, the pool model means you are not necessarily out of work.

Who qualifies

Required:

• MD or DO with a completed diagnostic radiology residency
• ABR certification or your jurisdiction's equivalent, or board-eligible with residency complete
• An active, unrestricted medical licence in your country of practice
• 2+ years reading independently after residency, currently or recently practising at high volume
• Precise command of structured reporting systems relevant to your subspecialty (BI-RADS, LI-RADS, Lung-RADS, TI-RADS and similar)
• Written and spoken English fluency
• At least 15 hours a week, with capacity to concentrate hours when a stream is time-boxed

Preferred: fellowship training (neuro, MSK, body, chest, breast or paediatric), US licensure and board certification, prior annotation or AI evaluation experience, teaching or rubric design, and published or sustained technical writing.

What it pays

$200–400 per hour, Mercor's published range and among the highest on this board. The ad does not say what places you within the band. Fellowship subspecialty training and US credentials are both listed as preferred, and they are the plausible levers, but that is our reading, not Mercor's statement. Assume the lower end until you see an offer.

Payment is weekly via Stripe or Wise for services rendered. Mercor cannot support H-1B or STEM OPT candidates. You are engaged as an independent contractor, so no benefits, and projects can be extended, shortened or ended early depending on needs and performance.

Worth knowing

Good:

• Top-tier hourly rate for non-clinical work
• Worldwide: no country restriction, and non-US certification is accepted
• Shared pool means access to several streams, not a single project
• Short tasks (as little as 10 minutes) fit around clinical schedules
• Weekly payment terms are published up front
• 124 hired this month, a sign of active demand

Less good:

• A 15-hour weekly minimum is a real commitment alongside a clinical job
• Band is wide, with no published basis for placement
• Throughput targets are set per stream and not disclosed in advance
• H-1B and STEM OPT holders excluded
• Contractor status: hours and duration are not guaranteed

About this listing

Posted by Mercor as an hourly remote contract open worldwide, confirmed open on 24 September 2026. Mercor states the work involves no access to confidential or proprietary information from any employer, client or institution, which also means studies or reports from your own practice are not to be brought in. See Mercor.