Luís is a seasoned Java developer with 20+ years of experience, currently at Azul Systems. He holds a Master's in Semantic Web with three international peer-reviewed papers, and is a member of the Jakarta Agentic AI specification team at the Eclipse Foundation. His work focuses on enterprise Java, Spring Boot, and MCP integrations.
devchallenges.tech lets developers learn by fixing broken code. I built it solo, vibe coded in Claude Code. The architecture: Jakarta EE 11 on Payara 7, challenges as YAML in Git, untrusted code sandboxed in-JVM and in Docker, user progress committed to GitHub instead of a database. Then production: Caddy, Docker Compose, push-to-deploy, and a Pix webhook paired with polling, idempotent on both paths. Plus where vibe coding cost me a day.
Jakarta Agentic AI is a new, vendor-neutral Jakarta EE specification for building AI agents. This session presents it and demonstrates it live on Payara Server, its reference implementation. You will learn the programming model: agents as CDI beans using @Trigger/@Decision/@Action/@Outcome/@HandleException over CDI events, plus an injectable LargeLanguageModel facade. Live samples swap the LLM provider (Ollama to Claude) in config, no code change