Modern AI systems need more than vector similarity to make sound decisions. To produce dependable results in complex enterprise environments, intelligent agents must understand what things are, what situational context they operate within, and how to execute procedural workflows.
This hands-on workshop introduces developers to three foundational graph paradigms that transform disconnected enterprise data into actionable AI context:
- Knowledge Graphs: Model real-world entities, taxonomy, and multi-hop relationships to ground LLM reasoning in verifiable facts.
- Context Graphs: Capture runtime dependencies, user session history, and environmental state to provide relevant context during inference.
- Procedural Graphs: Define step-by-step business logic, decision gates, and allowable execution paths for autonomous agents.
Through practical, hands-on exercises in Python, participants will construct connected graphs, execute Cypher graph queries, and integrate graph retrieval into RAG pipelines and agentic decision flows.
Key Takeaways
- Tri-Graph Architecture: Master the differences and synergies between knowledge, context, and procedural graphs.
- Graph Construction & Querying: Hands-on experience modeling nodes, edges, and properties for domain datasets.
- Graph-Augmented RAG (GraphRAG): Combine vector embeddings with structured graph traversals to eliminate LLM hallucinations.
- Procedural Agent Routing: Use graph traversal algorithms to guide multi-step agent actions safely and deterministically.
Target Audience
- Data engineers, backend developers, and AI engineers who want to advance beyond basic vector-only retrieval.
Prerequisites
- Basic familiarity with Python and fundamental understanding of data structures (lists, dictionaries, trees).

