Build and quality-check synthetic clinical data rooms — protocol, SDTM/ADaM datasets, TLFs, and patient narratives — so that every artifact reconciles end to end, then turn that same reconciliation work into prompts, golden outputs, and rubrics for scoring AI agents. Mercor lists $100–150/hr for this work at 20–25 hours per week minimum, US or Canada only.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work involves

You sit on both sides of the task: producing clinical data rooms and grading machines that try to produce or interrogate them. Generation work means constructing a coherent study package from the protocol outward — subject-level raw data, SDTM domains that pass conformance, derived ADaM, the TLF shells and outputs that draw from them, and narratives that reference the right subject at the right visit with the right AE onset. The realism bar is the point: a data room that looks plausible but where a discontinuation in DS never surfaces in the disposition table is a failed artifact, not a hard case.

The QC and authoring half is where the AI evaluation lives. You take a data room — synthetic, sometimes deliberately seeded with defects — and specify what a competent agent should find: which cross-file inconsistency, which traceability gap from CRF to submission dataset, which metric that folds incorrectly. That becomes a prompt, a golden output, and a rubric with enough granularity that two reviewers scoring the same agent response land in the same place. Expect to defend rubric line items when graders disagree, and to rewrite tasks that turned out to be ambiguous rather than hard.

What the screen looks for

Mercor's screening is AI-led and follow-up heavy. It probes for hands-on ownership rather than adjacency: EDC build through database lock, running or remediating SDTM conformance findings, sitting in the room when a lock slipped. Vocabulary alone does not survive the second question — expect to be asked which validator, which version, which rule numbers you argued about, and what you did when the sponsor and the CRO disagreed on a define.xml entry.

Logistics

Fully remote, US or Canada residency required
20–25 hrs/week minimum; 30+ strongly preferred and materially affects selection
Largely asynchronous with some synchronous calibration on rubric disputes
No degree requirement — years and depth of hands-on data management are the gate
$100–150/hr is the observed band on the platform, not a guarantee for any given engagement

Clinical Data Manager / CDISC

What the work involves

What the screen looks for

Logistics

Interview prep — 10 questions