Living Knowledge Graphs for Science

Session Track: AI Engineering

Session Time: 08:30 pm - 09:00 pm November 6

Session description

Saif Haobsh will demonstrate how Neo4j can turn static literature into a living discourse graph that branches, self-documents, and fuels agentic RAG workflows. You will learn how to: - Encode papers in a compact Q-C-E-I (Question-Claim-Evidence-Investigation) schema and load it as strongly typed nodes/edges - Compile any subgraph into an interactive narrative: the compiler consults a JSON widget-map, arranges pills, accordions, timelines, and evidence meters, and ships a lightweight HTML/CSS/JS bundle—no proprietary code exposed - Expose the graph to LLMs through RAG and Model-Context-Protocol adapters, yielding traceable answers instead of hallucinations - Apply the pattern to a life-science case study and export a narrative report that is fully back-linked to its Neo4j subgraph Attendees will leave with starter Cypher scripts, a widget-map template, and design heuristics for building graph-native AI systems that evolve safely, render themselves as rich user experiences, and surface insights text-only pipelines miss.