Join us for a technical deep dive into Mina, a solution to manage medical question-answer pairs across diverse healthcare domains. We’ll explore how to address LLM hallucination through grounding techniques, design Q&A data models in Neo4j with real-world considerations, and use Neo4j as both a vector store and a graph database. Additionally, we’ll dive into the architecture of retrieval-augmented generation (RAG) pipelines and discuss key evaluation metrics.

Perfect for AI practitioners, knowledge graph enthusiasts, and those tackling the challenges of reliable LLMs in specialized fields.

Guest: Quinten Rosseel
https://www.linkedin.com/in/quintenrosseel
Slides: https://pitch.com/v/mina—neo4j-live-9m7qrj

GraphRAG: https://graphrag.com/
Nomic AI: https://github.com/nomic-ai/nomic
Hugging Face MTEB: https://huggingface.co/spaces/mteb/leaderboard

0:00 – Introduction and Welcome
1:06 – Guest Introduction
4:05 – AI Agents and Their Role in Modern Workflows
13:27 – Taming LLM Hallucinations for Medical Q&A: Overview
20:08 – Demo: How the Medical Q&A System Works
29:58 – Understanding and Mitigating LLM Hallucinations
36:50 – Using Neo4j for Graph-Based Retrieval and Search Optimization
45:58 – Advanced Techniques: Graph RAG, Indexing, and Context Retrieval
54:30 – Evaluating AI Accuracy and Performance Metrics
1:06:02 – Closing Remarks and Upcoming Neo4j Events

#neo4j #graphrag #knowledgegraphs #genai #healtcare #chatbot