Sam Julien is director of product at CopilotKit and coauthor of Agentic GraphRAG, an O’Reilly book about combining knowledge graphs with AI agents. He builds infrastructure and interfaces that help developers turn enterprise AI into useful, trustworthy products.
From developer education to AI product leadership
Julien joined Auth0 in 2018 as a content engineer and later worked in developer relations at Auth0 and Okta. He taught Angular development, created a course on upgrading AngularJS applications, and helped organize the Angular Portland community. He also wrote Getting Started in Developer Relations and created the Developer Microskills newsletter. His approach to technical writing treats publishing while learning, seeking feedback, and practicing consistently as professional skills.
At Writer, which he joined in April 2024, Julien built the developer-relations program and supported launches involving AI Studio, Palmyra models, and tool calling. His responsibilities expanded into SDKs, documentation, and integrations before he became lead product manager for Writer’s IT and developer platform. He helped ship SDK 2.0 features including application retrieval, asynchronous jobs, knowledge-graph integration, and model delegation. His account of moving into product management connects developer advocacy with customer research, usability, positioning, and cross-functional product launches.
At CopilotKit, Julien works on open-source infrastructure connecting AI agents with applications, the AG-UI protocol, generative interfaces, and the transition from open-source adoption to enterprise products.
- Graph-based retrieval for dense enterprise knowledge. Vector search can confuse products, dates, and attributes when documents share specialized vocabulary or chunking separates facts from context. In his AI Engineer World’s Fair talk, Julien illustrated those failures with historical Apple products and similar mobile-phone specifications. Knowledge graphs preserve relationships that support multi-document reasoning and more transparent answers.
- Architecture shaped by operational reality. Julien described how Writer’s research and engineering teams addressed graph-database scaling, cost, and query limitations through context-aware splitting, graph-derived JSON stored in a Lucene-based search engine, and Fusion-in-Decoder techniques for combining evidence across passages. He presents these as collaborative engineering choices, not personal inventions.
- Agentic GraphRAG as production architecture. Coauthored with Anthony Alcaraz, Agentic GraphRAG extends graph-based retrieval into agent memory, reasoning, orchestration, evaluation, governance, and production operations.
- Generative interfaces with human oversight. Julien argues that increasingly capable agents need interfaces allowing people to inspect, approve, edit, and redirect consequential actions. He favors task-specific controls over expecting a chat window to handle every interaction.
- Product-minded engineering. His approach to engineering leadership brings developers into customer discovery, defines success before implementation, and uses instrumentation to measure whether shipped features improve outcomes.
