The work
Mercor is recruiting on behalf of an AI research lab building clinical AI for prior authorisation workflows. Day to day, you answer medication questions using supplied patient context, labelling, and source documents, then review model-generated responses against those same sources. When a response is wrong, incomplete, or unsafe, you correct it and write a structured annotation explaining why — the written rationale matters as much as the score.
Typical subject matter tracks what a busy PA or specialty pharmacy desk handles: medication history reconciliation, payer PA criteria, on- and off-label indications, dosing and renal adjustments, contraindications, interaction checks, and what documentation actually supports a request. A meaningful share of the job is knowing when to escalate rather than answer — thin evidence, conflicting guidance, missing clinical context, or a genuine safety flag.
What the screen looks for
- A real CPhT credential or equivalent hands-on experience, with specifics about setting, volume, and systems used.
- Prior authorisation fluency — payer criteria, appeals, step therapy, documentation requirements. Generic retail-counter experience is a weaker fit than specialty, PA, or medication access work.
- Evaluation judgment: can you distinguish a response that is wrong from one that is merely differently phrased, and can you hold a rubric consistently across dozens of items?
- Written clarity. Annotations are the deliverable. Vague feedback is a common failure mode.
Logistics
Fully remote and largely asynchronous, with scheduled calibration and feedback sessions you're expected to attend and act on. The commitment is 30–40 hours per week for the project duration — this is not a few-hours-on-the-weekend task queue. Pay has been observed at $35/hr; rates on Mercor vary by project and are set at offer, not guaranteed by this listing. Expect a screening interview, a scope-of-experience review, and often a short paid or unpaid calibration task before onboarding.