RAGFrameworks

Knowledge Graphs for RAG

by DeepLearning.AI × Neo4j

IntermediateCourseFree2 hours 4 minutes across 9 lessons, self-paced

Build a GraphRAG question-answering system over real SEC filings with Neo4j and LangChain.

Start LearningAdded Jul 4, 2026 · Updated Aug 2, 2026

Overview

Taught by Andreas Kollegger, who leads GenAI innovation at Neo4j, this two-hour-four-minute course runs nine lessons and builds its knowledge graph over real SEC filings rather than a toy dataset. It opens with knowledge graph fundamentals, then spends nineteen minutes on querying with Cypher, Neo4j's query language, before turning to RAG proper. Preparing Text for RAG covers getting text chunked and embedded so it can sit alongside graph structure. Constructing a Knowledge Graph from Text Documents builds the graph out of unstructured filings. Two further lessons, Adding Relationships to the SEC Knowledge Graph and Expanding the SEC Knowledge Graph, are the heart of it: the relationships between filings, companies and form types are exactly the structure a vector index discards. The longest lesson, Chatting with the Knowledge Graph at twenty-three minutes, assembles a question-answering system on Neo4j and LangChain that combines a vector index over text chunks with graph traversal, so the model receives connected context instead of a bag of similar paragraphs. Six of the nine lessons ship runnable code, ninety-nine of the course's one hundred and twenty-four minutes. On Coursera the same course carries a 4.8 out of 5 rating across 99 reviews with 10,538 learners enrolled, and DeepLearning.AI lists it free during its learning-platform beta.

At a Glance

Topic
RAG
Level
Intermediate
Format
Course
Cost
Free
Duration
2 hours 4 minutes across 9 lessons, self-paced
Provider
DeepLearning.AI × Neo4j
Hands-on
Yes — code/exercises
Certificate
None

What You’ll Learn

  • Write Cypher queries to create, manage and retrieve graph data in Neo4j
  • Model unstructured documents as nodes and relationships in a knowledge graph
  • Chunk and embed text so it sits alongside graph structure
  • Construct a knowledge graph from real SEC filings end to end
  • Add and expand relationships between companies, filings and form types
  • Combine a Neo4j vector index with graph traversal for hybrid retrieval
  • Build a question-answering chatbot on Neo4j and LangChain
  • Format graph query results into context an LLM can actually use

Highlights

  • Built and taught by Neo4j's GenAI innovation lead, not a generic course author
  • Uses real SEC filings, so the relationship modeling is a genuine problem rather than a contrived one
  • Six of nine lessons ship runnable code — 99 of the course's 124 minutes
  • Rated 4.8 out of 5 from 99 reviews on Coursera, with 10,538 learners enrolled
  • Free on DeepLearning.AI during its learning-platform beta

Who It’s For

Best For

  • RAG engineers whose vector-only retrieval breaks down on multi-hop questions
  • Developers with highly relational data such as filings, org charts or supply chains
  • Teams evaluating GraphRAG before committing to a graph database
  • LangChain users who want a retrieval layer beyond similarity search

Prerequisites

  • Basic Python — the lessons are notebook-driven
  • Familiarity with LangChain, or DeepLearning.AI's 'LangChain: Chat with Your Data' first
  • A working understanding of embeddings and vector similarity search
  • No prior Neo4j or Cypher experience needed; the course teaches Cypher from scratch

FAQ

What is Knowledge Graphs for RAG?

A short, code-heavy course from DeepLearning.AI and Neo4j on using graph structure, not just vector similarity, to retrieve context for an LLM. Across nine lessons you learn Cypher, build a knowledge graph from real SEC filings, and assemble a question-answering system on Neo4j and LangChain that traverses relationships a vector index throws away.

Is Knowledge Graphs for RAG free?

Knowledge Graphs for RAG is free to access.

What level is Knowledge Graphs for RAG for?

Knowledge Graphs for RAG is aimed at a intermediate audience. Recommended background: Basic Python — the lessons are notebook-driven, Familiarity with LangChain, or DeepLearning.AI's 'LangChain: Chat with Your Data' first, A working understanding of embeddings and vector similarity search, No prior Neo4j or Cypher experience needed; the course teaches Cypher from scratch.

How long does Knowledge Graphs for RAG take?

Expect roughly 2 hours 4 minutes across 9 lessons, self-paced. Most learners work through it at their own pace.

What will I learn from Knowledge Graphs for RAG?

You'll learn: Write Cypher queries to create, manage and retrieve graph data in Neo4j; Model unstructured documents as nodes and relationships in a knowledge graph; Chunk and embed text so it sits alongside graph structure; Construct a knowledge graph from real SEC filings end to end; Add and expand relationships between companies, filings and form types; Combine a Neo4j vector index with graph traversal for hybrid retrieval; Build a question-answering chatbot on Neo4j and LangChain; Format graph query results into context an LLM can actually use.

Topics

knowledge graphsGraphRAGNeo4jCypherLangChainRAG

Sources

This page was written from 3 sources, 1 on domains other than deeplearning.ai.

  1. 1.deeplearning.aiknowledge graphs ragvendor
  2. 2.coursera.orgknowledge graphs rag
  3. 3.deeplearning.aiknowledge graphs ragvendor