Alex Gilmore is a senior AI solutions architect at Neo4j who builds open-source infrastructure for turning disconnected enterprise information into usable context for AI applications. His projects include Neo4j Runway, Neo4j’s Model Context Protocol servers, and Neocarta, spanning knowledge-graph construction, persistent agent memory, and enterprise data discovery.
In 2024, as a Neo4j consulting engineer, Gilmore used Google Cloud, LangChain, and Neo4j to transform music-review transcripts into a searchable knowledge graph. He subsequently examined how graph data models shape retrieval, comparing document hierarchies, question-centered structures, and topic summaries. With Neo4j Runway, he translated those concerns into a Python library that analyzes relational data, proposes graph models, generates ingestion code, and loads the results into Neo4j.
At AI Engineer World’s Fair 2025, Gilmore demonstrated graph-backed agent memory connecting Claude Desktop to Neo4j through the Model Context Protocol. His implementation represented memories as typed entities, factual observations, and explicit relationships, producing inspectable knowledge that applications could update throughout a conversation and reuse across sessions or compatible clients. Neo4j’s role-based access controls also offered a practical mechanism for limiting access to sensitive information.
By 2026, Gilmore was developing Neocarta, an open-source system for building enterprise semantic layers over distributed data. His Google Cloud implementation connects BigQuery schemas, business terminology, catalog metadata, and query histories into a navigable graph, helping agents identify relevant tables, interpret columns, and infer useful joins.
- Retrieval through deliberate graph design: Different graph structures produce different retrieval behavior; useful context depends on modeling documents, questions, topics, and relationships appropriately.
- Persistent, inspectable agent memory: Explicit entities and relationships make conversational knowledge reusable, understandable, and accessible across compatible applications.
- Neo4j Runway: Language-model-assisted discovery and ingestion reduce the work required to convert relational datasets into functional knowledge graphs.
- Neocarta: Graph-based semantic layers connect technical metadata with business meaning, allowing agents to navigate enterprise information without loading an entire data estate into each prompt.
