Building Resilient AI Platforms: Financial Fraud Detection
summary
Blueprint for a multicloud, event‑driven microservices platform using Apache CFKA and Apache Flink with a dual feature store (offline + low‑latency online), model registry, and decision engine. Real numbers: P50 latency 47 ms, P99 89 ms, 45 TPS, 99.99% availability. Case study detects a network of 47 linked accounts across 12 institutions and prevents an estimated $2.3M loss. Includes graph features, behavioral biometrics, distributed training, sandboxed experiments, and federated learning for collaborative models.
AI-written summary · about
chapters (12)
- 0:00Introduction and Overview
- 1:58Engineering the Front Lines: Building Better Platforms
- 2:22Evolution of Fraud Detection Systems
- 3:49Architecting a Resilient AI-Powered System
- 7:40Real-Time Fraud Detection in Action
- 9:49Multi-Cloud Architecture for Resilience
- 11:54Data Engineering for Advanced Fraud Detection
- 20:43Federated Learning and Collaborative Defense
- 21:57Event-Driven Microservices and Biometric Integration
- 24:39Observability and Performance Optimization
- 26:05Business Impact and Future Technologies
- 28:24Conclusion: Winning the War Against Fraud
related talks
13mBuilding Financial Fortresses: Real-Time Fraud Detection
13mTransforming Financial Services with Real-Time Analytics
25mReal-Time Fraud Detection at Scale
16mOperationalizing AI in Finance: Real-Time Analytics
26mEngineering Excellence: Platforms and Architectures for Real-Time AI Analytics
46mData Analytics-Based Detection At Scale
10mReal-Time Fraud Prevention at Scale
51mReal-time Fraud Detection: Challenges and Solutions