What the work involves

Micro1 recruits practicing and formerly practicing attorneys to evaluate how large language models handle compliance questions. On a given project you might be writing realistic prompts a compliance officer would actually send ("can we onboard this PEP-adjacent entity under our current risk rating?"), grading two model responses against a rubric, or rewriting a flawed answer into a reference response that a senior reviewer would sign off on. Much of the value you add is catching outputs that read fluently but are wrong in ways only a specialist notices — citing a superseded FinCEN rule, conflating GDPR and CCPA obligations, treating guidance as binding, or giving a confident answer where the honest answer is jurisdiction-dependent.

Projects vary. Some are heavy on adversarial red-teaming (can you get the model to draft something that facilitates structuring or sanctions evasion?), some on long-document review (policy manuals, SAR narratives, disclosure filings), some on rubric-based scoring at volume. You will be asked to justify scores in writing, and your written rationale matters as much as the score itself — it is training signal, not paperwork.

What the screen looks for

  • A verifiable compliance track record. JD plus real regulatory work: in-house compliance, a financial-institution advisory practice, enforcement-side experience, or privacy counsel work. Bar admission is typically expected, active status often not required.
  • Depth under follow-up. The AI interviewer will push past your first answer — expect it to ask which specific regulation, which safe harbor, what changed after a particular enforcement action.
  • Calibration. Strong evaluators distinguish "legally wrong" from "technically defensible but bad practice" from "I would need the client's jurisdiction to answer." Overconfidence scores poorly.
  • Clear, terse writing. Rationales are read by non-lawyers.

Logistics

Fully remote and largely asynchronous, with occasional calibration calls scheduled against US business hours. Contributors typically commit 10–20 hours per week; some projects support more, and volume fluctuates as engagements start and end. Pay is hourly and observed in the $90–150/hr range for this specialization — the actual rate depends on project, seniority, and how you assess in calibration, and is set by micro1, not guaranteed here. Engagements are 1099/contractor, and conflict-of-interest and confidentiality terms apply.