An AI lab is training a language model on how finance professionals build and narrate slide decks, and needs senior FP&A analysts to write, grade, and correct that output. Work covers monthly reporting slides, bridge and waterfall exhibits, headcount and opex pages, scenario decks, and the commentary that goes with them — observed at $70/hour, 5–20 hours per week, fully remote and async.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work actually is

You will spend most sessions on one of three tasks. First, authoring: given a prompt resembling a real finance request — "build the Q3 opex bridge for the business review" — you produce the slide, the chart structure, and the commentary a controller would actually sign off on. Second, evaluation: two model-generated versions of a reporting page land in front of you and you decide which is better and write out why, in terms specific enough that an engineer who has never built a variance bridge can act on it. Third, correction: you take a model output that is nearly right and rewrite it, flagging exactly where the logic, the labelling, or the narrative broke.

The finance content matters more than the visual polish, but both are graded. A waterfall where the bars reconcile but the bridge categories double-count price and volume is a failed task. So is a headcount exhibit that is arithmetically fine and says nothing a CFO would care about. The model needs to learn the difference between a slide that presents numbers and a slide that makes an argument.

What the platform screens for

Ethos runs an AI voice screen before any task access. It probes for concrete FP&A tenure — which company, what scale of P&L, which cycles you owned — and then goes a layer deeper on method: how you actually construct a price/volume/mix bridge, what belongs in a monthly reporting pack versus an ad hoc deck, how you handle a variance you cannot explain by the deadline. Expect follow-ups that test whether your answers come from doing the work or from describing it.

Depth is tested by follow-up, not by the first answer
Written samples or a described deck structure carry more weight than job titles
Evaluation judgment — can you say why one version is better — is screened separately from domain skill

Logistics

Fully remote, fully asynchronous; no standing meetings
5–20 hours per week, self-scheduled, with more available to reviewers who perform well
$70/hour is the observed rate on this listing, not a guarantee; rates vary by task type and tenure on the platform
Work is batched — periods of steady volume alternate with quieter stretches

Senior Financial Analyst (FP&A)

What the work actually is

What the platform screens for

Logistics

Interview prep — 10 questions