Agentic Knowledge Graph Construction
by DeepLearning.AI and Neo4j
Use Google's ADK to build agents that turn CSV files and product reviews into one Neo4j knowledge graph.
Overview
Agentic Knowledge Graph Construction is a DeepLearning.AI short course built with Neo4j and taught by Andreas Kollegger, who works on generative AI at Neo4j. It was announced on August 27, 2025, and the course page lists a last update of July 17, 2026. The page lists 3 hours 8 minutes across twelve lessons and a graded quiz. It opens with What is a Knowledge Graph? (10 minutes) and the architecture of the multi-agent system (9 minutes), then spends 42 minutes introducing Google's Agent Development Kit across two parts. The remaining lessons follow the pipeline: Understanding User Intent (17 minutes), File Suggestions (12 minutes), Schema Proposal for Structured Data (29 minutes), Schema Proposal for Unstructured Data (16 minutes) and Knowledge Graph Construction in two parts (46 minutes). The running example looks for the root cause of product issues by combining CSV product and supplier data with unstructured product reviews. Agents propose nodes such as products, parts and suppliers, along with relationships such as a product containing a part or a part provided by a supplier. Review chunks are mined for the products and issues they mention. The graphs built from those plans are connected into one knowledge graph. A companion MIT-licensed repository, neo4j-contrib/agentic-kg, holds the working multi-agent system. Enrollment is free during the DeepLearning.AI learning platform beta, which the page describes as a limited-time offer.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Course
- Cost
- Freemium
- Duration
- 3h 8m listed, self-paced (the individual lesson times add up to about 3h 15m)
- Provider
- DeepLearning.AI and Neo4j
- Hands-on
- Yes — code/exercises
- Certificate
- Available
What You’ll Learn
- ✓Explain how a knowledge graph extends RAG systems that store chunks in vectors
- ✓Build specialized agents with Google's Agent Development Kit across two code lessons
- ✓Design an agent that establishes the user's goal for the graph before extraction
- ✓Have an agent suggest which input files should feed the knowledge graph
- ✓Generate node and relationship schema proposals from CSV product and supplier data
- ✓Propose what to extract from unstructured product review chunks, including products and issues
- ✓Construct graphs from the agents' plans and connect them into one knowledge graph
Highlights
- •Structured and unstructured sources share one pipeline, with 45 minutes of lessons on schema proposal across the two data types.
- •The companion repo neo4j-contrib/agentic-kg (53 stars on November 9, 2026) has separate intent, file suggestion, schema proposal and graph construction agents under one coordinator.
- •The repo calls itself 'a reference implementation for learning and experimentation, not a production tool' and lists unstructured import, evaluations and more GraphRAG tools as open work.
- •A learner commenting on the launch post reported that the model often skipped the final get_proposed_construction_plan call, so inspect the tool-call sequence when you run the lessons.
- •The graded assignment and the course accomplishment require a DeepLearning.AI PRO membership, while the lessons are free to enroll in during the platform beta.
Who It’s For
Best For
- ✓RAG engineers whose retrieval needs explicit relationships between entities
- ✓Developers learning multi-agent design with Google's Agent Development Kit
- ✓Teams joining product, parts or supplier records with unstructured customer feedback
Prerequisites
- •Python familiarity is required for the code lessons
- •Experience with RAG systems is recommended by the course page
- •Basic Cypher query knowledge helps but is not required
FAQ
What is Agentic Knowledge Graph Construction?
A short intermediate course for Python developers with RAG experience who want agents to plan and build knowledge graphs. You use Google's Agent Development Kit and Neo4j to create a multi-agent system that proposes graph schemas from structured and unstructured data, then constructs and connects the resulting graphs.
Is Agentic Knowledge Graph Construction free?
Agentic Knowledge Graph Construction offers free content, with paid options for certificates or premium features.
What level is Agentic Knowledge Graph Construction for?
Agentic Knowledge Graph Construction is aimed at a intermediate audience. Recommended background: Python familiarity is required for the code lessons, Experience with RAG systems is recommended by the course page, Basic Cypher query knowledge helps but is not required.
How long does Agentic Knowledge Graph Construction take?
Expect roughly 3h 8m listed, self-paced (the individual lesson times add up to about 3h 15m). Most learners work through it at their own pace.
What will I learn from Agentic Knowledge Graph Construction?
You'll learn: Explain how a knowledge graph extends RAG systems that store chunks in vectors; Build specialized agents with Google's Agent Development Kit across two code lessons; Design an agent that establishes the user's goal for the graph before extraction; Have an agent suggest which input files should feed the knowledge graph; Generate node and relationship schema proposals from CSV product and supplier data; Propose what to extract from unstructured product review chunks, including products and issues; Construct graphs from the agents' plans and connect them into one knowledge graph.
Topics
Sources
This page was written from 3 sources, 2 on domains other than deeplearning.ai.