Most AI agent systems fail not because the models are weak, but because they lack an operational foundation. This session explains AgentOps around two core pillars: MLflow as the system of record for agent lifecycle management, and an AI Gateway as the control plane for routing, governance, and cost management. We will explore how to treat agents as versioned, testable AI assets, tracking prompts, tools, evaluation datasets, and performance metrics in MLflow to create reproducibility and monitoring.
You’ll learn how an AI Gateway enables model-agnostic routing, guardrails, rate limiting, payload logging, and observability, providing the governance layer required for enterprise deployment. By the end of the session, attendees will understand how to design a production-ready AgentOps architecture that combines experimentation, evaluation, monitoring, and control, turning agentic systems into managed, measurable, and scalable applications.
Technical Level of Session: Technical practitioner
Supported by
