Build a Real-Time Fraud-Detection Stream for a Card Issuer
Overview
What this challenge is about.
Design the stream topology: authorization events in, customer-feature state (30-day rolling) maintained in state store, scoring function applied per event, fraud-score emitted to a downstream Kafka topic with sub-80 ms end-to-end latency. Implement using Apache Flink (or Kafka Streams). Use a pre-trained scoring model (provided in ONNX format) wrapped in a Java/Scala wrapper for low-latency inference. Benchmark at 600 events/sec sustained, 1200 burst. Deliver source code, 10-page writeup, and latency benchmark report at p50/p95/p99/p99.9.
The Brief
What you'll do, and what you'll demonstrate.
Build a Kafka + Flink stream-processing pipeline that scores card authorizations in under 80 ms p99 at 600 events/sec sustained.
Earning criteria — what you'll demonstrate
- Design a stream topology with state stores for real-time feature computation
- Apply Flink (or Kafka Streams) under hard latency budgets
- Wrap an ONNX model for low-latency JVM inference safely
- Benchmark stream latency at the tail, not just the mean
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Career mappings coming soon.