What the work looks like
You receive AI-produced artifacts — a patient education handout, a payer-facing utilization summary, a clinical trial results deck, a budget-impact spreadsheet — and grade them against a rubric supplied by the client. Grading is not a thumbs up or down: you mark specific spans, name the failure (fabricated citation, wrong dosing units, misread inclusion criteria, chart axis that misrepresents effect size), and write a short justification a non-clinician reviewer can follow. Much of the value you add is catching errors that read as fluent and confident, which is where general-purpose annotators consistently miss.
Expect a mix of formats. Slides matter more here than candidates usually assume — you are judged on whether you can spot a mislabeled Kaplan-Meier curve, an unreadable table, or a deck that buries the safety signal, so real facility with PowerPoint and Google Slides is a hard requirement rather than a line item.
What the screen measures
- Verifiable clinical depth. Five-plus years in a healthcare or clinical role — practice, research, regulatory, pharma medical affairs, health economics, nursing, or clinical operations all qualify. The AI interviewer will push on your actual scope of work, not your titles.
- Written precision. Feedback quality is the product. Vague or hedged critique gets filtered early.
- Rubric discipline. Whether you can grade to a stated standard instead of substituting your own clinical preferences.
- Honest scoping. Saying "outside my area" when a task drifts into a specialty you don't hold is scored positively, not against you.
Logistics
Fully remote, fully async, hourly. Work arrives in batches and volume fluctuates by project; most contributors treat this as a 5–20 hour/week supplement rather than a full load. No set shifts, but tasks often carry turnaround windows measured in days. An advanced degree (MD, PhD, MPH, MSN, PharmD, MS) is preferred and tends to correlate with placement at the upper end of the observed band.