AI is reshaping Java development from two directions at once, and most developers are only paying attention to one of them. This presentation uses both.

This session showcases a Quarkus microservices application where user-submitted content is evaluated in real time by an LLM. The AI evaluation service consumes Kafka events, uses the Quarkus LangChain4j extension to generate vector embeddings, and autonomously approves or rejects submissions via cosine similarity — all within a fault-tolerant, observable pipeline.

And here's the twist: I will live code the application with AI. Using Claude Code, custom Skills, and spec-driven development, a structured approach built around product specs, coding standards, and incremental feature tasks, the application will be implemented with agentic collaboration. I'll also demonstrate other agentic tools including Codex, Gemini CLI, Copilot CLI, Cursor, Kiro, and Warp.

You'll leave with concrete Quarkus + LangChain4j integration patterns, a practical framework for structuring agentic development, and the confidence to apply both in your next Java project.