Code 2025
Context Engineering: Connecting the Dots with Graphs — Stephen Chin, Neo4j
Overview
This talk explores context engineering as a method to enhance AI applications by moving beyond simple prompt engineering. It emphasizes the importance of providing AI models with dynamic, curated, and structured context to improve signal over noise, leading to more relevant and reliable outputs. The core thesis is that by treating AI development as information architecture, engineers can achieve superior results and gain greater control over AI behavior.
Who should watch
- AI Engineers
- Product Managers
- Builders working with LLMs
- Those facing challenges with AI context windows and attention focus
- Developers seeking to improve retrieval patterns and agentic memory
Key takeaways
- Context engineering evolves prompt engineering by providing dynamic and wider scopes of information to AI models, leading to better results.
- Effective context engineering involves integrating prompt design, retrieval augmented generation (RAG), and managing both short-term and long-term AI memory.
- Knowledge graphs offer a structured representation of information, connecting entities through relationships, which is highly beneficial for AI by providing organized context.
- Graph RAG enhances traditional RAG by incorporating knowledge graphs into the retrieval process, enabling more relevant results by leveraging relationships and community groupings.
- Knowledge graphs provide explainable AI by allowing visualization of data and interactions, aiding in performance analysis and identifying areas for improvement.
- An open-source tool, the knowledge graph builder, can ingest documents and use an LLM to construct a knowledge graph, facilitating structured data for AI.
- Graph memory retrieval uses graphs to store and query information, combining vector embeddings with graph algorithms for richer context and memory for AI agents.
- Agentic multi-step retrieval using tools like the Neo4j MCP server allows LLMs to traverse and query knowledge graphs, yielding highly detailed and specific information.
Notable quotes
*We've become slaves to um AI programming, to prompt development, to building things off AI models.*
*This allows us to think not like prompt engineers but like information architects where we're building the model context which actually gives us superior results coming out of the AI.*
*LMS are only as good as the quality of the response that they're getting from the data. So if you give them bad data if you get them garbage then you are going to get garbage back out again.*
Unofficial community note. Prefer the recording for nuance.