GraphRAG & Knowledge Graphs with Neo4j: Build AI Agents

About this course

This course contains the use of artificial intelligence.

I am Mariano Flynn, and I personally designed this curriculum. I wrote and reviewed every script and every class, and I used AI tools only as production support under my direction, with a full review of each class before upload.

Many AI assistants can answer simple questions, but they struggle when knowledge is scattered, connected, or changing. Naive vector RAG retrieves similar text chunks, yet it often misses relationships, context, and the reasoning path behind an answer. This course shows you a stronger architecture: combine the precision of vector search with the structure of a knowledge graph.

You will start by modelling data in Neo4j and writing Cypher, then build entity extraction pipelines that turn documents into nodes and relationships. After creating a traditional vector RAG baseline, you will measure its limits and use those results as a benchmark for everything that follows.

Next, you will implement GraphRAG step by step: community detection, local search for focused questions, global search for broad summaries, and hybrid retrieval that blends graph traversal with vector similarity. You will compare RAG and GraphRAG, learn when each approach wins, and choose the right architecture for production workloads.

The course then moves to agents. You will build persistent agent memory on a knowledge graph, create multi-hop reasoning agents, and use text-to-Cypher with guardrails to keep queries safe and reliable. You will also expose Neo4j through MCP so other tools and agents can use your graph safely.

In the final project, you will wire everything together into a complete LLM, GraphRAG, and agent application, and evaluate it against your baseline. By the end, you will be able to design, build, and evaluate production-ready retrieval and agent systems that truly know your data.

Prices, discounts and availability are set by Udemy. We may earn a commission when you purchase through links on this site.