GraphRAG for Law: Building Legal Reasoning Agents with Neo4j and LLMs

Session Track: AI Engineering

Session Time: 03:30 am - 04:00 am November 6

Session description

In legal AI, trust is everything and hallucinations by language models can lead to serious consequences. To build reliable systems, developers need more than raw LLM output. They need structured, source-grounded reasoning. In this talk, Aleksandr Khazov, founder of LEXRAG, will show how GraphRAG architecture—powered by Neo4j, APOC, and native vector search—can unlock the full potential of graph-native retrieval over a complete legislative codebase. He will walk you through the process of modeling granular legal entities as graph nodes, enriching them with embeddings, and orchestrating hybrid retrieval that combines semantic search with deep legal structure.

Speaker

Aleksandr Khazov

CEO and CTO, LEXRAG

Aleksandr Khazov combines more than 20 years of legal experience with technical leadership to build cutting-edge LegalTech solutions. As founder of LEXRAG, he leverages GraphRAG architecture with Neo4j at its core to enable precise retrieval and reasoning over complex legal texts. His work integrates structured legal ontologies, vector embeddings, and LLMs to power legal research assistants. Aleksandr personally designed and deployed the entire pipeline, from graph modeling to production inference. His dual background in law and engineering positions him at the forefront of applying graph-native AI to real-world legal challenges.