What the work involves

AfterQuery contracts actuaries to build evaluation and training data for frontier AI labs. On a typical task you receive a prompt and one or more model responses — a loss reserving walkthrough, a rate indication, a mortality assumption discussion, an IFRS 17 or LDTI question — and you judge whether the reasoning is technically sound, whether the numbers reconcile, and whether the approach would survive review under actuarial standards of practice. You then write the correction or the reference answer: a clean demonstration of how the triangle should have been developed, how the trend selection should have been justified, or why the model's chain ladder assumption fails on that data.

Ranking tasks ask you to order responses against each other and explain the ordering in writing. The written rationale matters as much as the ranking — it is the training signal. Expect to spend more time articulating why a Bornhuetter-Ferguson selection was inappropriate than on the arithmetic itself.

What the platform screens for

  • Credential verification. SOA or CAS designation (ASA, FSA, ACAS, FCAS), or substantial exam progress with three or more exams passed.
  • Practice depth under follow-up. The screen probes a stated specialty — if you claim P&C reserving, expect questions on tail factors, case reserve adequacy, and diagnostics. Breadth claims across life, health, P&C, and pensions get tested at the seams.
  • Evaluation judgment. Whether you can distinguish a genuine methodological error from a defensible alternative selection, and whether you know when to flag ambiguity rather than force a verdict.
  • Written clarity. Every task output is prose. Reviewers read for precision and for reasoning that a non-actuary annotator could follow.

Logistics

Fully async — no standing meetings, no shift coverage. Work is claimed from a queue and volume fluctuates with lab demand, so this suits people who can absorb an irregular pipeline rather than those needing guaranteed weekly hours. Most contributors work in blocks of a few hours. Pay is hourly within an observed $80–150/hr band, generally tracking credential level and specialty scarcity; it is not guaranteed and rates for specific projects are set at assignment.