How Capitec uses graph databases to protect its clients

Fraud investigations already look like a graph. Chains, clusters, and multi-hop links. This session shows how Capitec runs a graph feature pipeline that scores more than 3.5 million records daily in about 2 hours, then routes results into case management. Derick Schmidt, head of product, and Jan Ehlers, data scientist and data engineer, break down how they turned investigation maps into a repeatable pipeline for Authorized Push Payment fraud detection.

What you’ll learn

• Why APP fraud creates network patterns that tables miss
• How graph features feed a standard ML workflow
• Production flow from graph to scoring and case management
• Leakage prevention and sampling strategy