Java GenAI libraries are getting more mature every day. It is now incredibly easy to spin up a Spring AI "Hello World" example and connect your code to models like GPT or Gemini with just a few lines of configuration. While these "magical" setups are excellent for quick prototypes, they often hide away the control and reliability needed for real-world business needs. This session addresses the gap between those basic demos and enterprise-friendly implementations. We will look at how to move past the default configurations and take full control of the execution flow. The goal is to ensure that it's your application and not the LLM in the driver's seat, allowing for better observability and predictable behavior when the AI interacts with your internal data and tools. Finally, we will tackle the essential requirements for production: Security, Identity, and Human-In-The-Loop (HITL) workflows. We will explore how to secure the Model Context Protocol (MCP) so your tools are identity-aware, and discuss strategies to validate risky operations, ensuring that high-stakes actions are verified by a human before execution.

Talk Level:
INTERMEDIATE

Bio:
Panche is a Principal Software Engineer at G+D Netcetera and the current jug.mk leader. An experienced Java developer by trade, he has a passion for IoT, 3D printing, and recently, Generative AI. He is known as the creator of pulse.eco, a core initiator of netaville.mk, and an active contributor to the Macedonian IT community.