Learn with Neo4j's New "Get to Know Graph & GenAI" Webinar Series >>
Session Track: AI Engineering
Session Time:
Session description
LLMs struggle to reason on large graphs. Recent advances in agents provide LLMs external toolings, yet there is no such agent that can choose and execute a variety of graph algorithms for complex reasoning. In this session, we will present a graph algorithms agent (MCP server) that can execute all algorithms in the GDS library on Neo4j. We will give a live interactive demo, allowing the audience to ask any graph-algorithmic questions in natural language and get answers. The server can be taken off-the-shelf and used on any Neo4j database with GDS. We will also go through the implementation behind the GDS-agent as well as challenges and future opportunities for community.
Senior Software Engineer, Neo4j
Brian has been working on graph algorithms, graph machine learning, and retrieval-and-reasoning on graphs with LLMs. His recent work includes finding ways of improving retrieval-augmented generation on graphs and building agents that reason over graphs by using graph algorithms as tools.