Learn with Neo4j's New "Get to Know Graph & GenAI" Webinar Series >>
Session Track: Knowledge Graphs
Session Time:
Session description
As GraphRAG becomes a popular method for grounding LLMs in enterprise knowledge, graph developers and data architects are increasingly asked to make their knowledge graphs “AI-ready.” Yet most graphs, especially those built through automatic or ad hoc methods, contain hidden quality issues that undermine LLM performance. These issues often go unnoticed until they manifest as vague answers, poor retrieval, or untraceable hallucinations. In this hands-on workshop, Panos Alexopoulos will guide graph practitioners, data modelers, and AI engineers through the process of assessing and improving knowledge graph quality for GraphRAG pipelines. Using real-world examples, you will learn how semantic structure, lexical clarity, and modeling consistency influence grounding outcomes, and how to spot the silent failures that disrupt them. By the end, you will leave with a reusable knowledge graph evaluation checklist, a deeper understanding of semantic design trade-offs, and strategies for proactively improving the reliability of your LLM-augmented systems.
Semantic Data and AI Solutions Lead, Triply BV
Panos Alexopoulos is a data semantics and AI practitioner with 20 years of experience designing, evaluating, and deploying knowledge graph solutions across industry and academia. He is the author of "Semantic Modeling for Data: Avoiding Pitfalls and Breaking Dilemmas" (O'Reilly, 2020), and he currently leads Semantic Data and AI Solutions at Triply BV, in Amsterdam, Netherlands. Panos is also a seasoned educator, delivering training programs on topics like knowledge graphs and large language models.