What the work actually looks like

You'll spend most of your time reading model output and deciding whether it would hold up in a real encounter. A task might present a clinical vignette — a 62-year-old with new-onset atrial fibrillation, a toddler with a febrile seizure, a patient asking about tapering gabapentin — alongside one or two AI-generated responses. You judge the differential, the workup sequence, the drug choice and dose, the follow-up interval, and whether the answer stayed inside the scope an NP would actually practice within. Much of the volume is head-to-head comparison: two responses, pick the better one, and write the rationale that explains why in terms a non-clinician annotator could not have produced.

The written rationale is the deliverable, not the rating. Reviewers who write "incorrect dosing" get flagged; reviewers who write "metformin started at 1000 mg BID without renal function noted — should be 500 mg daily titrated, and contraindicated below eGFR 30" are the ones kept on the project. Patient education artifacts get a second axis: reading level, actionability, and whether the material quietly overstates certainty.

What the screen is looking for

Mercor's screening is AI-led and moves fast. It verifies the licensure and years claim, then probes clinical depth with follow-up questions that go one layer past your first answer — expect to be asked why that antibiotic, what changes if the patient is pregnant, what would make you refer out. It also tests calibration: whether you can distinguish an answer that is wrong from one that is merely not how you would have phrased it. Confident wrongness and vague hedging both score poorly. Written English is assessed directly from your typed responses, so treat them as work samples.

Logistics

  • Fully remote, asynchronous task queues; no scheduled shifts or patient contact
  • 20+ hours per week is a real floor, not a target — throughput is tracked
  • Duration roughly 4–6 weeks, immediate start
  • Payment begins task-based; approval of your first task within the stated window converts you to hourly for the remainder
  • Pay band reflects observed rates for this listing and varies by specialty and task type; it is not guaranteed