Agentic Fraud Investigation with TigerGraph

For the TigerGraph Hacker House Goa challenge, an Agentic Fraud Investigation solution was developed using TigerGraph Savanna and a custom Python-based agent. The system analyzes fraud cases by examining card history, device data, and billing regions, mimicking a human analyst. Utilizing the HHGOA_IEEE dataset containing 590,742 transactions, the agent evaluates entity relationships to generate structured evidence. It determines if sufficient information exists to act or if customer verification is required to resolve the investigation.