Your question is Fraud Detection for GEICO Claims. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
Given a dataset describing customer claims (some are known to be fraudulent, the rest are not), how would you approach the problem of building a model to identify which future claims are likely fraudulent?
Asked in the technical interview stage. Discuss how you would define labels, address class imbalance, prevent leakage, select and validate models, and operate the system safely in production.
Address data preparation, model serving, evaluation, monitoring, retraining, and the business tradeoffs between missed fraud and manual-investigation volume.