Expert · 5–6 weeks
Fraud Detection with Rules + ML Hybrid
Real-time transaction scoring: rule engine (velocity, geo mismatch) + ML model score (Drools + exported model or external API). Explainable decline reasons and analyst review queue.
Why this project
Shows AI beyond chatbots — hybrid systems are how fraud actually works in banks and fintech.
Tech stack
Java 17Spring Boot 3DroolsPython ML model via API or DJLKafkaRedis
Skills demonstrated
- Rules engine
- ML integration
- Real-time scoring
- Explainability
Interview talking points
- Rules vs ML division
- Feature engineering examples
- Explainability for analysts
- Handling model drift
Problem statement
Real-time transaction scoring: rule engine (velocity, geo mismatch) + ML model score (Drools + exported model or external API). Explainable decline reasons and analyst review queue.