What the work involves

You are given model outputs and asked to say, with reasons, whether a clinician would let them stand. Typical items include a 15-year-old describing restriction and compensatory exercise, a parent asking whether their daughter's weight loss warrants a referral, a question about refeeding syndrome monitoring, or a request for "healthy fasting" advice from an account that reads adolescent. You rate against a rubric — clinical accuracy, safety escalation, tone — and then write the part that actually matters: a short, specific critique naming what is wrong and what the correct answer would have been. Some batches ask you to rewrite the response as a gold-standard reference; others ask you to compare two model outputs and defend a preference.

The recurring failure modes are domain-specific. Models give calorie numbers and BMI cutoffs to users who should not receive them, miss atypical anorexia in a normal-weight teen, treat ARFID as picky eating, offer weight-loss framing to a restricting patient, and either under-escalate genuine medical instability or over-escalate every mention of food into an emergency-room referral. Distinguishing those two errors reliably is most of the job.

What the platform screens for

micro1's screen is AI-led and conversational, roughly 20–40 minutes, and it follows up. Expect to be asked how you would handle a specific vignette, then pressed on why — the follow-up is where credential claims either hold or collapse. Screeners look for a verifiable clinical background in adolescent eating disorders (child and adolescent psychiatry, adolescent medicine, ED-specialised psychology, dietetics with a CEDS or equivalent caseload), familiarity with DSM-5-TR criteria and medical instability guidelines, and the ability to state a judgment in two sentences rather than five paragraphs of hedging.

Logistics

  • Fully remote and asynchronous; you claim batches when they open rather than holding fixed shifts.
  • Volume is irregular — busy weeks then nothing. Most reviewers treat this as a supplement to clinical work, at 5–15 hours a week.
  • Rate is set at onboarding based on credential and calibration performance; the $45–95/hr band is what has been observed, not a guarantee.
  • Expect an unpaid or nominally paid calibration set before live work, and periodic re-calibration against other reviewers.