GraphSentinel- Agentic fraud investigation

GraphSentinel and JEVelric are agentic fraud investigation systems developed for the TigerGraph challenge. These systems automate the complete workflow, starting from risk alerts or customer reports to produce traceable records. By utilizing a knowledge graph, they gather evidence, evaluate model belief, and recommend actions based on specific policy citations. JEVelric specifically utilizes the IEEE-CIS/Vesta dataset, processing approximately 590,000 card transactions and identity signals to mimic human analyst behavior through a structured trigger-to-action architecture.