What the work involves

You receive AI-generated artifacts — a three-statement model in Excel, a purchase price allocation memo, a variance analysis deck, a revenue recognition write-up — and assess them the way a reviewing manager would. That means checking whether the numbers tie, whether the accounting treatment is defensible under the relevant framework, whether assumptions are labeled and internally consistent, and whether the presentation would survive a partner or CFO review. You then write structured feedback: what's wrong, why it's wrong, and what the correct output would have looked like.

Much of the grading is rubric-driven, so a large part of the job is calibration — applying the same standard across dozens of samples rather than drifting toward leniency or nitpicking. Some tasks ask for a numeric score across dimensions (factual accuracy, analytical rigor, formatting, completeness); others ask for a side-by-side preference between two model outputs with a written rationale. Slide-deck evaluation is a real component here, not an afterthought: layout, chart choice, label accuracy, and narrative flow are all scored.

What the platform screens for

  • Verifiable depth. Expect follow-ups that push past résumé language into specifics: which standards you've applied, what you actually built, where you've seen practitioners get it wrong.
  • Error-spotting under time pressure. Screens often include a flawed artifact and ask what you'd flag first.
  • Judgment on ambiguity. Many AI outputs are not wrong so much as under-supported. Reviewers who can articulate that distinction score well.
  • Written clarity. Feedback is the deliverable; terse or vague comments don't clear the bar.

Logistics

Fully remote and asynchronous. Work is drawn from a task queue, typically in blocks of a few hours, with volume that fluctuates by project. Most contributors treat this as 5–20 hours per week alongside existing work. Pay is hourly at rates observed in the $80–120 range — actual offers vary by specialization, credentials, and project, and nothing here is guaranteed.