What the work involves

You will spend most of your sessions reading inpatient documentation and judging it against what a practicing hospitalist would actually write. That means reviewing H&Ps, daily progress notes, and discharge summaries for accuracy and completeness, then evaluating AI-generated versions of the same artifacts — flagging omissions (the pending culture nobody carried forward), hallucinations (a medication or comorbidity that appears from nowhere), clinical inaccuracies, and documentation gaps that would matter to the next clinician on service or to a coder. Feedback is structured and written: annotations against a detailed guideline, plus a rationale an engineer who has never rounded on a patient can act on.

The second half of the job is guideline work. Annotation rubrics for clinical documentation break down fast at the edges — how to score a plausible-but-unsupported assessment, when a terse note is appropriately terse, how to treat a summary that is accurate but omits the discharge contingency. You will be asked to raise ambiguity rather than paper over it, and to help define edge cases so the next reviewer scores them the same way you did.

What the platform screens for

  • Current inpatient practice. Mercor's screen probes recent, hands-on note-writing — census size, service setting, which artifacts you personally author. Teaching-only or purely outpatient backgrounds tend not to clear this.
  • MD with 2+ years post-residency, US-based, in hospital medicine or internal medicine.
  • Calibration under follow-up. Expect to be pushed on a judgment call: why an omission is clinically material rather than merely stylistic, and where you would defer to a guideline you disagree with.
  • Written precision. The deliverable is prose an engineering team reads. Vague clinical impressions do not survive.

Logistics

Remote, part-time, asynchronous, built around an active clinical schedule — evenings, post-call days, and off-service weeks. Volume tends to arrive in batches tied to model releases, so expect uneven weeks rather than a fixed commitment. Familiarity with Epic and prior AI/ML annotation or model-evaluation experience are preferred, not required. The $170/hr figure reflects rates observed on this listing and is not a guarantee.