What the work involves
You receive AI-generated artifacts — a draft response to a regulator's information request, a risk-assessment matrix in Sheets, a board-level compliance update deck — and assess them the way you would assess a junior associate's work before it left the firm. That means checking whether cited rules and regulations actually say what the model claims, whether the tone and hedging are appropriate for a supervisory audience, whether control descriptions map to real testing evidence, and whether the formatting and structure would survive internal review. Feedback is written, structured, and tied to a rubric: what is wrong, why it matters, and what a correct version looks like.
Expect a mix of scoring, side-by-side comparisons of two model outputs, and occasional rewrites or gold-standard authoring. Subject matter tends to cluster around AML/BSA and sanctions programs, SEC/FINRA and banking-regulator examinations, consumer protection rules, policy and procedure drafting, remediation plans, and regulatory correspondence.
What the screening looks for
- Verifiable depth. The AI interview probes specifics: which regulators you've faced, what kind of requests you've responded to, how you handled a finding. Vague answers get follow-ups.
- Evaluation judgment. Can you separate a stylistic preference from a substantive error, and can you catch confident-sounding regulatory citations that don't exist?
- Written clarity. Feedback that a model trainer can act on, not a redline with no rationale.
- Slide and spreadsheet fluency. The listing calls this out explicitly; a lot of the artifacts are decks.
Logistics
Fully remote, fully asynchronous, contractor engagement with no minimum commitment. Most experts treat this as 5–15 hours per week alongside primary work, though volume fluctuates with project demand — periods of steady tasking alternate with quiet stretches. Pay is hourly at rates observed in the $80–120 band; actual offers depend on domain, assessed depth, and project budget, and are set by the platform rather than guaranteed.