From MLOps to AI Platforms, Evaluation, Governance, and Agents in Production

Machine learning infrastructure is changing quickly as teams move from classic model deployment towards broader AI platform work that includes evaluation pipelines, governance controls, observability, and support for foundation model based workflows. As production systems become more complex, the challenge is no longer only how to serve models reliably, but how to build platforms that support repeatable, accountable, and scalable AI delivery.

This session will examine how modern organisations are rethinking ML platform design in response to these demands. It will look at the shift from narrow MLOps practices towards platform approaches that unify experimentation, deployment, monitoring, reproducibility, and operational oversight. Particular attention will be given to the growing importance of evaluation frameworks, workflow orchestration, and platform level controls for agent based and generative systems.

The talk will also address the organisational side of platform maturity, including how engineering teams balance speed with governance, reduce friction between research and production, and create shared standards that make AI systems safer and more dependable over time. Drawing on real production challenges, the session will offer a grounded view of what it takes to operate machine learning and AI systems responsibly at scale.

This talk will cover

• The shift from classic MLOps to broader AI platform engineering
• Evaluation pipelines for machine learning and generative systems
• Reproducibility, monitoring, and operational controls in production
• Governance patterns for high impact AI workflows
• Supporting agent based systems without losing reliability or oversight