What the work actually involves

You build accounting problems hard enough to break a model, then judge whether the model's answer holds up. A typical task starts from a fact pattern you construct — a multiple-element arrangement under ASC 606, a sale-leaseback with a variable payment, a step acquisition changing consolidation conclusions, a Level 3 fair value estimate — and requires you to supply a defensible answer with citations to the relevant codification guidance. Other tasks run the other direction: you receive a model's technical memo, footnote draft, or 10-Q disclosure and mark where it is wrong, where it is right for the wrong reason, and where it is technically correct but would never survive review by an auditor or the SEC's Division of Corporation Finance.

The distinction between those last two categories is most of the value you add. Models produce fluent, well-formatted accounting prose that cites the correct ASC topic and still lands on a conclusion no controller would sign. Feedback that says only "incorrect" is not useful; feedback that identifies the specific judgment the model skipped — the missing performance-obligation analysis, the discount rate it never justified, the disclosure the guidance requires that it omitted — is what the training pipeline consumes.

What the screen looks for

  • Concrete reporting history: which filings you touched, whether you prepared or reviewed, public or private, and what the hardest technical issue in your recent work actually was.
  • Codification fluency under follow-up. Expect the screener to push a scenario one step past the textbook answer and see whether you reason or recite.
  • Willingness to write. Graders who explain their reasoning in a paragraph are worth more than graders who tick boxes, and the screen tests for that in your answer length and structure.
  • Honest availability. Weekly hours you can genuinely commit, with tasks pulled from a queue rather than assigned in meetings.

Logistics

Fully remote and asynchronous — no standing calls, no client hours. Contributors commonly work in blocks of a few hours, often 5–15 hours a week, though volume moves with active projects and is not guaranteed. Pay is hourly within the observed $75–150 band; the upper end tends to go to people with SEC reporting or Big 4 technical accounting depth and, frequently, a CPA. Calibration feedback arrives in writing, and continued task flow depends on your work surviving review.