What the work actually involves
You'll spend most of your time in an annotation interface, not an EHR. Tasks arrive as batches: a clinical encounter's source documentation paired with an AI-generated note, summary, or code assignment. Your job is to judge whether the output holds up against real documentation and coding standards — does the note support the code, is specificity lost, has the model invented a diagnosis the record doesn't substantiate, has it dropped a comorbidity that changes the picture. You then write the finding up so a machine learning engineer with no coding background understands what went wrong and why it matters.
A second stream of work is guideline contribution. Annotation guidelines for clinical documentation are written by people who are not coders, and they break in predictable places — query-worthy versus codeable, clinical validation versus coding validity, when unspecified is genuinely correct. Coders who flag those ambiguities early, with a concrete example attached, tend to get pulled into guideline review rounds and higher-rate work.
What the screen looks for
Mercor's screening is AI-led and follow-up heavy. It verifies the certification and that you are currently practising — a CPC from 2016 with no recent coding work is a common rejection. Expect probes into specificity, MEAT/TAMPER-style documentation support, HCC logic if you claim it, and at least one exchange conducted in or about your second language, since C1+ proficiency in speaking, listening, and writing is a stated requirement rather than a nice-to-have. It also checks whether you can articulate disagreement in writing rather than just marking something wrong.
Logistics
- Remote, US-based, part-time — minimum 10 hours per week
- Asynchronous with flexible scheduling; no fixed shifts, though batches have turnaround windows
- Observed pay band is $45–65/hr, varying with certification stack, language, and task complexity — not guaranteed
- Work volume is project-driven and can be uneven between cohorts