tbc

AI agents are moving fast from demos to production, but the gap between a clever prototype and a reliable system is wide. This session takes a technical deep-dive into how AI agents actually work under the hood, using Google Cloud's Gemini Enterprise Agent Platform as the reference architecture. We'll cover core building blocks like the Agent Development Kit (ADK) for code-first agent logic and Agent Runtime for scaling long-running, stateful agents backed by Memory Bank for persistent context, how Agent Identity, Agent Registry, and Agent Gateway provide centralized governance and enterprise-grade guardrails, and the failure modes that show up once you leave the sandbox: hallucinated tool calls, context drift, and runaway loops, including how Agent Simulation, Agent Evaluation, and Agent Observability provide execution traces and real-time visibility into agent reasoning. Attendees will leave with practical patterns for building agents that are observable, controllable, and safe to deploy at enterprise scale.