System design interview question
Design a Real-Time Fraud Detection System
A real-time fraud detection system inspects every transaction as it happens (typically a card payment, login, or money transfer) and decides — within ~100 ms — whether to allow, challenge, or block. Underneath it combines a rules engine for deterministic policies (block this card BIN; require 3DS over $500), a feature-engineering pipeline that turns the raw event stream into hundreds of behavioural and velocity features, and a low-latency ML model server that returns a risk score. The system must operate at thousands of TPS, learn continuously from analyst feedback, and minimise false positives — every wrong block is a lost sale and an angry customer.
Practice this question step by step with an AI interviewer that reviews your requirements, your architecture and your trade-offs, and scores the design.