What the work actually looks like

You are handed clinical scenarios — a presenting complaint, a history, some vitals and labs — alongside model-generated responses that propose a differential, a workup, or a management plan. Your job is to decide whether the reasoning is sound, whether the differential is appropriately broad without being padded, whether the next steps match current guidance, and whether anything in the output could plausibly harm a patient if acted on. Most tasks are structured: a rubric, a rating, and a written justification. The justification is the product. Reviewers who write "incorrect, would order CBC" get filtered out fast; reviewers who write "the response anchors on GERD and never addresses the exertional component, which makes an ACS rule-out mandatory before any PPI trial" get more work.

Expect a mix of task types across a project: side-by-side comparisons of two model answers, single-response grading against a rubric, free-text rewrites of flawed clinical explanations, and adversarial prompt writing where you construct cases designed to expose a model's weak spots — atypical presentations, polypharmacy, geriatric syndromes, the patient whose chief complaint isn't the actual problem.

What the screen is looking for

micro1 runs an AI-led interview before any human review. It tests three things: that your clinical background is real and specific, that you can reason aloud under follow-up rather than recite a guideline, and that you can articulate why an answer is wrong in a way another physician would accept. Prior annotation, RLHF, or model-evaluation experience is listed as preferred and does help, but clinical depth carries more weight — the platform can teach rubrics faster than it can teach medicine.

Logistics

  • Contractor, remote, US-based; 1099 arrangement, no benefits.
  • Fully asynchronous. Tasks are pulled from a queue when you have time; there are no shifts.
  • Volume fluctuates by project. Many physicians treat this as 5–15 hours a week alongside clinical practice; some projects open with higher short-term capacity.
  • The $126–441/hr band reflects observed rates across micro1 medical projects and varies with specialty, task complexity, and calibration performance. It is not a guaranteed rate, and hours are not guaranteed.