Most Graph-RAG demonstrations fall flat. They dive into complex code, use simplistic unrealistic fraud scenarios, or miss the real hidden challenge: detecting fraudulent patterns across disconnected transactional systems with no foreign keys between them.
So I decided to do something different. Let me introduce you to Slick Eddie. He's a fraudster working with four buddies to steal gift card funds across multiple uncoupled systems.
Here's the thing: I wanted to catch Eddie using only ChatGPT-like prompts. No code. But I hit a wall. The fraud pattern and data volume were too complex for prompting alone.
That's when I decided to partner with Oracle 26ai and an AI Coding Assistant. Together, we built a multi-phase detection pipeline in PL/SQL and PGQL using some clues, fraud fingerprint embedding, similarity search to find suspicious transaction bursts, Property Graphs to reduce 3.7 million paths to 132, and balance validation to isolate the exact fraud sequence.
This presentation reveals the real story. You'll see what worked, what didn't, and how your curiosity combined with an AI assistant can solve problems in days that would otherwise take months.
Ready to catch Slick Eddie?
9:30 am
