What the work actually is

You receive a scenario packet: a trial balance, a trader's PnL summary, a term sheet, a chain of emails, maybe a purchase agreement and a stack of aging reports. Some of it matters and some of it was planted to see whether you notice it doesn't. Your job is either to produce the deliverable — a reconciliation that ties, a linked pro-forma model, a covenant compliance schedule, a profit-pool waterfall — or to review an AI-generated attempt and say precisely where it departs from what a controller or partner would accept. Reviews are not scores in a vacuum; you write the justification, cite the source document, and state what the correct treatment would have been.

The subject matter is deliberately non-markets. Expect ASC 805 deferred-revenue haircuts and PPA amortization phasing, ASC 606 timing questions, month-end mark and commission adjustments, PO / P-card / three-way-match controls, capital accounts, and debt covenant definitions that differ from GAAP by a few contractually specified add-backs. A CFA and a portfolio-management history are not what this funnel is screening for.

What the platform screens for

  • Verifiable hands-on years in product control, financial control, FP&A, technical accounting, treasury/AP–P2P, commercial-banking credit, or professional-services finance — described at the level of what you closed, reconciled, or signed.
  • Depth under follow-up. The screen will push on a treatment you name until you either cite the standard and the mechanics or run out. Broad familiarity reads as thin quickly.
  • Excel craft you can describe structurally: how your 3-statement model balances, where circularity lives, how you prove a reconciliation.
  • Evaluation judgment — can you separate a defensible-but-different assumption from an actual error, and rank errors by materiality rather than counting them.
  • Writing. CFO/partner-ready narrative is an explicit deliverable, and grading rationales are read by other reviewers.

Logistics

Fully remote and asynchronous, with task batches rather than shifts. Most contributors work 10–20 hours a week around a primary role; throughput and consistent quality matter more than which hours you choose. Onboarding typically includes a paid or unpaid calibration task on a representative scenario, and continued work depends on agreement with reviewer consensus. Pay is stated as observed on this listing, not guaranteed, and can vary by task type and complexity.