What the work actually involves
You build evaluation scenarios — multi-step research tasks with verifiable answers — that a frontier AI agent is then asked to solve. In practice that means picking a question hard enough to be interesting (a five-year revenue segment reconstruction from 10-Ks, a patent family's forward-citation chain, a player's split statistics under a specific condition), locating the authoritative source that settles it, and writing the task plus a defensible gold answer with the source trail attached. You will also rewrite and edit tasks other contributors wrote, and document model failures granularly enough that an engineer can act on them — not "the model was wrong" but which step it skipped, which source it invented, and what the correct value is.
What the platform screens for
AfterQuery's screening is less about your CV headline and more about whether you can prove things. Expect to be pushed on where a number comes from, whether your source is primary or a secondary aggregator, and how you'd handle a case where two credible sources disagree. Domain expertise in at least one of the listed areas — sports analytics, financial analysis, patent research, academic/admissions research, business intelligence — is preferred and tends to be the differentiator, but the non-negotiables are English writing quality, source discipline, and the ability to work unsupervised against a formatting spec you'll be held to.
Logistics
- Independent contractor, fully remote, async — no fixed shifts.
- 10–20 hours per week recommended; described as an extended contract, compatible with other commitments.
- $15 is the rate observed on the listing; the period was not confirmed by the platform, and "top-of-market" is the platform's own framing, not a guarantee.
- Expect an onboarding period where early submissions are reviewed closely and returned with revision notes before your throughput ramps up.