Design real-time fraud detection (Stripe Radar) + DSA review

Interview questions to prep

1. How do you avoid duplicate combinations without a set — what's the index-passing trick?
2. Compare with combination-sum-ii where each candidate can only be used once.

Interview questions to prep

1. Walk me through designing Stripe Radar — real-time fraud detection.
2. How do you balance false positives (blocked good users) vs false negatives (fraud)?

Interview questions to prep

1. What kinds of features (velocity, graph, behavioral) actually catch fraud?
2. How do you serve graph-based fraud features in real-time without exploding compute?

Interview questions to prep

1. Design a system that detects fake or fraudulent contributed content in maps reviews, photos, or edits.
2. How would you combine user reputation, content features, graph signals, velocity features, and human review?
3. What metrics balance user trust, false positives on legitimate contributors, and time-to-removal for abuse?

Interview questions to prep

1. Fraud labels arrive weeks late (chargebacks). How do you train and monitor under delayed labels?
2. How would you use early proxy signals (e.g., risk score, manual review) when chargebacks are slow?