Agent Interaction Graphs: Evaluating Multi-Agent Systems with Graph-Based Reasoning

Session Track: Graph Memory & Agents

Session Time: 09:30 am - 10:00 am April 15

Session description

Multi-agent systems fail for a multitude of reasons from agent-to-agent communication, tool use, hallucination, and complex multi-turn user conversations. Logs and traces are necessary but not sufficient as conversations create a needle-in-a-haystack problem and no automatic root cause analysis.

In this session, we model agent executions as an interaction graph in Neo4j and use this knowledge graph, attach our evaluations and run graph queries to pinpoint critical issues, recurring failure points and bottlenecks based on deep contextual relevancy from a graph. Turns agent evaluation from a spreadsheet into a navigable graph.

Speaker

Vincent Koc

Futurist | Lecturer | Artificial Intelligence, Machine Learning and Data

Vincent Koc is a highly accomplished, commercially-focused engineering technologist with a wealth of experience in data-driven disciplines. He holds a fellowship at the Institute of Managers and Leaders Australia, where he serves as a thought leader and mentor to the next generation of data professionals. With over a two decades of experience in the field, Vincent has worked across a wide range of industries, including finance, telecommunications, travel, and fast-moving consumer goods. He has driven data-driven projects for major organizations such as Qantas, Telstra, Volkswagen, Expedia, the Australian Federal Government, and the NSW Government. Previously Vincent has held critical positions at major Australian organisations.