Jeremy Adams-Casañas is a senior developer advocate at Neo4j working on graph-grounded AI agents, trustworthy retrieval, and controlled software development. His career spans infrastructure automation at Puppet, container security at Twistlock, developer platforms at GitHub, and software-delivery workflows at Dagger.
At Puppet, he wrote about integrations with VMware and Docker; at Twistlock, he worked on partnerships and alliances. He later led Dagger’s ecosystem team and helped build its partner program, bringing experience in developer adoption to the emerging problem of AI-generated code overwhelming established review and delivery practices.
At AI Engineer World’s Fair 2025, Adams-Casañas and Kyle Penfound demonstrated how coding agents can inherit an engineering organization’s existing tools, containerized environments, automated tests, and GitHub workflows. Adams-Casañas emphasized software-delivery guardrails: generated changes should pass through the same validation and isolation mechanisms that govern human-written code. Typed inputs and outputs, narrowly scoped tools, and portable execution environments make agent behavior easier to constrain and inspect.
After joining Neo4j, he extended that operational focus into graph-grounded AI agents and the relationships that give models more useful context.
- GraphRAG with relationship-aware retrieval. Adams-Casañas combines vector similarity search with graph traversal so agents retrieve relevant text alongside connected concepts, documents, and sources. He has demonstrated Neo4j Aura workflows that express retrieval logic through Cypher and low-code agent configuration.
- Graph modeling through concrete examples. In an explanation of the card game Dobble, he uses finite projective planes and graph relationships to show why any two cards share exactly one symbol.
