LLM + Graph Database for Retrieval Augmented Generation (RAG)

ABSTRACT: LLMs are often like the know-it-all at a bar - they can quickly and confidently produce realistic sounding answers to just about any question - even if the answers are complete fabrications. But an LLM can be grounded in reality by combining it with a Knowledge Graph in order to prevent hallucinations, and to prevent unauthorized access to sensitive data. This presentation will show you the benefits of Graph Databases over regular databases and how to use GenAI with RAG to eliminate hallucinations, enforce security, and improve accuracy. We will also discuss why a vector index plus Knowledge graph provides better, smarter, faster results than a pure vector database. We will demonstrate an end-to-end retrieval pipeline. The code in the demo will be available in a Jupyter notebook on Github for you to reuse. SPEAKER BIOs Soham Dhodapkar is a Solution Engineer at Neo4j, helping users all over the world solve complex problems using the power of Graph Databases. He is an AI and Data Science enthusiast with research experience in Machine Learning, Analytics and Natural Language Processing. Soham is a conversationalist, likes stargazing, hiking and is a recreational tennis player. https://www.linkedin.com/in/sohamdhodapkar/ Andreas Kollegger is a technological humanist. Starting at NASA, Andreas designed systems from scratch to support science missions. Then in Zambia, he built medical informatics systems to apply technology for social good. Now with Neo4j, he is democratizing graph databases to validate and extend our intuitions about how the world works. Everything is connected. You can find presentation details here: * Andreas's slides: https://drive.google.com/file/d/1LYA8xgIbz5-SGn5rhMvrbFvd1-e4GeUW * Soham’s Github demo: https://github.com/neo4j-partners/neo4j-generative-ai-aws * Content Library: https://drive.google.com/drive/folders/1R7VHC_U09XXF3a1wD7kJp3SyoOYdAvTS More resources: * Free Graph Academy Courses: https://graphacademy.neo4j.com/ * Neo4j with Generative AI models using Langchain * Build a Neo4j-backed Chatbot using Python * and others * Start building with AuraDB free: https://neo4j.com/cloud/aura-free/ * Stay informed with the Developer Newsletter: https://neo4j.com/tag/twin4j/ https://www.meetup.com/sf-bay-acm/events/297992915/ 0:00 Chapter Intro 4:51 Presentation & Speaker Intro 6:44 Presentation 7:30 The GenAI Stack, Andreas Kollegger 7:56 Generative AI 13:13 How do we integrate? 13:22 Generative AI is a new layer in the Stack 15:50 Retrieval Augmented Generation (RAG) 21:18 3) Pure Data Retrieval 40:07 2) Mixed Text & Data Retrieval 42:29 Knowledge Graphs 51:57 Making it Real (Examples) 59:29 Thanks! 1:00:37 Grounding Your LLM, Soham Dhodapkar 1:03:35 Demo: SEC EDGAR Filings 1:15:20 RAG with Vectors & Graph 1:25:54 neo4J auraDB 1:26:20 GraphAcademy