Building Financial Fortresses: Real-Time Fraud Detection
summary
Claims real subsecond detection using Kafka Streams and a feature store, microservices with autoscaling, circuit breakers and rule‑based fallbacks; deploys with automated canary analysis to a replica environment using synthetic transactions. Reports concrete outcomes: 78% reduction in detection time, 60–63% drop in false positives, sustained 99.99% availability and 42% cost savings. Observability tracks inference latency, prediction confidence, data drift and gives explainability reason codes for every decision.
AI-written summary · about
chapters (11)
- 0:00Introduction and Speaker Background
- 0:37The Challenge of Real-Time Fraud Detection
- 2:29Transition to Streaming Architecture
- 3:39Resiliency Patterns for AI Systems
- 4:40Specialized Observability for ML Systems
- 6:00Service Level Objectives for Fraud Detection
- 7:11Chaos Engineering for Fraud Systems
- 8:12Technical and Business Gains from SRE Principles
- 9:08Balancing Compliance and Innovation
- 10:07Blueprint for AI Fraud Detection Systems
- 11:14Conclusion and Final Thoughts
related talks
13mBuild Real-Time Analytics Platforms for Financial Crime Detection
13mTransforming Financial Services with Real-Time Analytics
51mReal-time Fraud Detection: Challenges and Solutions
47mReal-time Fraud Detection Challenges and Solutions
29mBuilding Resilient AI Platforms: Financial Fraud Detection
25mReal-Time Fraud Detection at Scale
27mArchitecting Sub-Second Analytics Pipelines
29mReal Time Fraud Detection with Stateful Functions