Insurance Specialist

$60–80/hrrate as listed

Write and grade insurance reasoning tasks — underwriting, claims, and risk assessment — for a frontier AI lab's training pipeline. It's a W-2 role through Cincinnatus LLC at roughly 35 hours/week on weekdays, with observed pay in the $60–80/hr range.

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

What the work involves

You'll spend most of your time on two activities: authoring insurance problems hard enough that a strong model gets them wrong, and scoring model responses against structured rubrics. A task might be a commercial property submission with ambiguous COPE data, a bad-faith exposure question on a first-party claim, or a reserving judgment call where the technically correct answer diverges from what a junior adjuster would do. You write the reference solution and the reasoning trail behind it — not just the conclusion.

The grading side is where most reviewers get filtered out. You'll read a model's answer, decide whether its reasoning is sound or merely arrives at the right number by accident, and write feedback specific enough for a researcher with no insurance background to act on. "Wrong" is not useful; "applied occurrence-based trigger logic to a claims-made policy, then compounded the error in the tail coverage discussion" is. You'll also help refine the rubrics themselves as failure patterns emerge, and reconcile scoring differences with other SMEs so the training signal stays consistent.

What the platform screens for

Depth of line-specific expertise. Mercor's AI interview will pick one area you claim — P&C underwriting, life actuarial, claims litigation management, reinsurance — and push several layers past the surface. Generalist answers stall out quickly.
Verifiable career progression. The listing is explicit: 8+ years at a recognized carrier, broker, or reinsurer, with visible advancement. Titles, employers, and lines of business should match your resume exactly.
Prior rubric-based AI evaluation. This is stated as mandatory. Be ready to name the platform or program, describe the rubric dimensions you scored on, and explain a call where you disagreed with a fellow annotator.
Availability that matches the ask. 35 hours/week on weekdays is a near-full-time commitment; screens probe for conflicting engagements.

Logistics

Remote and largely asynchronous, though the weekday requirement means this doesn't stack easily with another full-time role. Employment is W-2 through Cincinnatus LLC as employer of record, with placement onto an AI lab's extended workforce — payroll, benefits, and compliance run through Cincinnatus. Pay in the $60–80/hr band reflects what candidates have reported for this listing; actual offers vary with specialty and screening outcome.

What they look for

8+ years in insurance underwriting, claims, actuarial, or risk management at a recognized carrier, broker, or reinsurer
Hands-on experience evaluating LLM outputs against structured rubrics or scoring criteria
Demonstrable career progression within a specialty line (e.g., Underwriter → Senior Underwriter → VP)
Ability to write precise, defensible technical rationale for non-specialist readers
35 hours/week weekday availability
Eligibility for W-2 employment in the United States

Before you apply

Would you pass this screen?

Drawn from what this listing actually requires. Answered in your browser — nothing is sent to us, and we do not store your answers.

Do you have 8 or more years of professional experience in insurance roles such as underwriting, claims, actuarial, or risk management at a recognized organization?
Insurance industry experience · required
Do you have hands-on experience evaluating LLM or AI model outputs against rubrics or structured scoring criteria?
AI/LLM evaluation experience · required
Can you reliably commit to at least 35 hours per week during weekdays for the duration of the contract?
Weekly availability · required

0 of 3 answered

Included with this listing

Interview prep — 11 questions

Matched to what this platform's screening actually measures. Answer each out loud in 60–90 seconds: claim → concrete example → implication.

What it tests: Tests whether the candidate has genuinely lived underwriting judgment, not just processed claims; reveals the depth of domain reasoning the AI lab will rely on to build training data.
A strong answer: The answer names a concrete underwriting scenario (property, liability, life, or specialty line), identifies 2–3 material risk factors that drove the decision, explains the trade-offs between acceptance and pricing, and shows how experience sharpened that judgment—not a generic 'I learned to be thorough.'
Where people lose it: Describing a decision so vague or procedural ('I checked the checklist') that it's indistinguishable from what a junior analyst would do; or claiming expertise in a line (e.g., reinsurance) without being able to articulate real underwriting tension.