2026-02-15

project 2: the fraud detection sliding window

project 2: the “fraud detection” sliding window

scenario: “we lost $50k in a chargeback scam.” you need to detect users who make >5 transactions in 10 seconds. the mission: write a stateful stream processor.

• tech: faust (python streaming) or pyspark streaming, redis.
• challenge: implement a “sliding window” aggregation.
• dev to prod:
  1. read from the project 1 kafka topic.
  2. use redis to store a rolling window of transaction counts per user_id.
  3. trigger an alert (print to console/slack webhook) if threshold > 5.
  4. prod requirement: handle “out of order” events (events arriving late).