AI Training and Evaluation Expert (Insurance Underwriting & Claims)

Weekday is an AI-native recruiting platform backed by Y Combinator and General Catalyst. We have built the largest white-collar talent database in India and have developed tools for running outbound recruiting campaigns.

We are currently running an AI training and evaluation project as a YC-backed AI data lab. We are seeking insurance experts to design and evaluate tasks that advanced AI models will be tested on.

About the Role

Advanced AI models are increasingly being applied to insurance tasks such as underwriting decisions, risk assessment, and claims adjudication. The effectiveness of these models depends heavily on the quality of the tasks created and evaluated by human experts.

Your role will involve creating complex underwriting and claims tasks that AI models should be able to handle but often fail to. You will evaluate the AI's performance, identifying where it is accurate, subtly incorrect, or confidently wrong. Precision is key—vague scenarios or lazy evaluations are not acceptable.

This is not repetitive back-office processing or data entry. Each task requires real insurance judgment, akin to the decisions a senior underwriter or claims assessor would make and defend. If you prefer structured, templated work, this role is not for you. However, if you enjoy dissecting complex risks or contested claims and articulating why a decision is flawed, this role will likely suit you.

Requirements

What You’ll Do:

• Design intricate underwriting scenarios, including applications, risk profiles, and eligibility or pricing calls, with clear and defensible model answers
• Construct claims cases that hinge on coverage determination, policy interpretation, exclusions, or fraud indicators, along with appropriate adjudication
• Review AI-generated underwriting and claims decisions line by line, grading them on accuracy, reasoning, and completeness
• Identify misinterpreted policy wording, missed exclusions, incorrect risk assessments, and decisions that seem plausible but are flawed
• Write precise rationales for evaluations to ensure the model and team learn from the feedback

Every task is reviewed, and you will be evaluated based on the quality and rigor of your submissions, not on the volume of tasks completed.

What We’re Looking For:

• 3+ years of hands-on experience in underwriting, risk assessment, or claims—across life, health, general, or specialty insurance lines
• Deep understanding of how risk is assessed and priced, and how claims are ultimately decided
• Strong ability to read and interpret policy wordings, including conditions, exclusions, and endorsements, and apply this knowledge practically
• Precision in writing—clearly articulating what is wrong and why, without unnecessary padding
• Low tolerance for misleading or nonsensical claims, whether from AI or others
• Reliability in meeting deadlines—task-based work requires timely completion
• Curiosity about how AI will transform insurance work, with a preference for actively shaping its development over passive observation

What You Get

• Compensation of ₹20,000–25,000 per task, paid for the depth of your analysis rather than hours worked
• Remote and flexible work that can be integrated with a full-time role
• A unique opportunity to observe and influence how frontier AI models are trained and evaluated in insurance, contributing to their improvement
• A fast-moving, direct team backed by Y Combinator
• Additional tasks and larger projects if your work consistently demonstrates high quality