REX: Self-learning Real-time Fraud Detection System
summary
A real-time credit-card fraud detection system built on Kafka, Akka-HTTP and Akka-Streams that achieves sub-20 ms latencies at several hundred transactions per second. It uses an auto-learning XGBoost model that continuously retrains via Kafka topics and deploys without downtime, with synchronous feature ingestion and controlled replication to ensure consistency.
AI-written summary · about
chapters (12)
- 0:00REX fraud detection
- 2:50ML key terms
- 5:20REX architecture
- 8:58Scoring node operation
- 12:07Performance optimization
- 15:26Fault handling & databases
- 18:40Real-time reference updates
- 20:03Fraud model improvement
- 23:57Scala in REX
- 28:55Next iteration
- 31:04gRPC integration
- 33:13Q&A
related talks
13mBuilding Financial Fortresses: Real-Time Fraud Detection
25mReal-Time Fraud Detection at Scale
13mTransforming Financial Services with Real-Time Analytics
47mReal-time Fraud Detection Challenges and Solutions
51mReal-time Fraud Detection: Challenges and Solutions
20mA Reactive Fraud Monitoring Engine for Instant Payments
10mReal-Time Fraud Prevention at Scale
13mBuild Real-Time Analytics Platforms for Financial Crime Detection