What the work actually is

You write the problems and you write the answers. A typical task starts with you constructing a realistic actuarial scenario — a rate indication for a commercial auto book, a reserve review with a distorting large-loss year, a mortality assumption update following an experience study — complete with the data, context, and ambiguity a working actuary would face. You then produce a "golden" reference response at the level an experienced professional would hand to a manager: method selected and justified, calculations shown, assumptions documented, conclusion defensible.

The second half is grading. You score model-generated responses against structured rubrics covering methodological soundness, calculation accuracy, assumption justification, and communication quality. The valuable part is not the score; it is the written rationale explaining why a chain ladder was wrong for a book with changing case reserve adequacy, or why a trend selection was unsupported by the data provided. Researchers read that prose. Vague grading is the main reason contributors are dropped.

What the screen looks for

  • Two or more years of hands-on actuarial work in any practice area — life, health, P&C, pensions, or ERM. Credentials (ASA, FSA, ACAS, FCAS, CERA, MAAA) and exam progress are bonuses, not gates.
  • Whether you can pick a method and defend it under follow-up, rather than recite a list of methods. Expect the screener to push back on your first answer.
  • Clear separation between actuarial analysis, underwriting decisions, and formal financial or legal advice — this boundary matters to the lab and gets probed directly.
  • Writing that stands on its own asynchronously, since nearly all feedback is read, not discussed.

Logistics

Fully remote and async. Minimum 20 hours per week, with 40+ preferred; work is drawn from a queue rather than scheduled. The one synchronous commitment is onboarding office hours and periodic calibration sessions, where graders align on rubric interpretation. Applications are reviewed on a rolling basis and the role starts immediately. Tooling proficiency (R, Python, SQL, SAS, Prophet, Emblem, ResQ, Radar) and rate-filing or statement-of-opinion experience strengthen an application but are not required.