William Lyon works on graph intelligence at Neo4j and writes about building applications with connected data. He is the author of Full Stack GraphQL Applications, a practical guide to a graph-backed web application.[1][2]
Start with Full Stack GraphQL Applications →
X · lyonwj Career
Lyon previously worked as a software developer at startups building mobile applications, quantitative finance tools and predictive APIs. A conference biography identifies his master’s degree in computer science from the University of Montana. He also worked on developer experience at Hypermode before his current Neo4j graph-intelligence work. His public projects combine database design with the application interfaces that expose it.[3][4][1]
Contributions
An application built around a graph
Full Stack GraphQL Applications teaches a complete web application using GraphQL, React, Apollo and Neo4j. It connects a backend graph model with frontend queries, then adds authentication, authorization, pagination and cloud deployment. The book’s purpose is to show how the layers fit together, so a developer can follow a data request through an application rather than learn each library in isolation.[2]
Memory that preserves relationships
Lyon’s current project list includes Neo4j Agent Memory and a memory-graph workshop. The memory system stores conversation history alongside extracted entities, relationships and reasoning traces. Its graph structure supports retrieving connected context, while vector search handles similarity. His workshop combines Neo4j and Pydantic AI in a news chat agent, giving developers a smaller application in which to inspect how memory affects a response.[1][5]
Tools, writing & talks
Writing, research & talks
Start here
Full Stack GraphQL Applicationsmanning.comBuilds a graph-backed web application from its API and React client through authentication and deployment.
Begin with the book if you know web development but are new to GraphQL or graph databases. Its end-to-end application provides the foundations for his more specialized work on agent memory.
Author[2]Neo4j's Will Lyon argues retrieval alone leaves agents amnesiac. He presents a context graph for agent memory and a method to distill grounded, governed skills as typed execution graphs, demoed with a patient-intake skill.
Catalog workOriginal source
Tools
- Neo4j Agent Memory
github.comA graph-native system for conversation, knowledge and procedural memory in AI agents.