Deploying AI in healthcare is difficult; deploying AI at scale is even harder. Over the past year, Vizient confronted a unique challenge: how do you operationalize more than 2,000 custom healthcare models—each tailored to different patient groups—without creating 2,000 deployment pipelines, 2,000 monitoring stacks, and 2,000 points of failure?

This talk reveals how we engineered a unified architecture that collapsed thousands of bespoke models into a single deployable system, enabling reliable, governed, and scalable delivery across hospitals and care settings nationwide. Attendees will learn the technical patterns, MLOps foundations, and architectural decisions that made this possible.

We’ll walk through:

• The strategy behind unifying 2,000 predictive models into one scalable deployment artifact
• How standardized MLOps, CI/CD, and platform engineering enabled massive model consolidation
• Key design choices for reliability, governance, lineage, and monitoring across highly variable clinical use cases
• The blueprint for extending this platform into agentic AI, including how emerging agents can safely operate across clinical workflows
• Lessons learned from scaling healthcare AI beyond traditional model deployment patterns

By the end of this session, participants will understand how to architect AI systems—not just models—that scale across thousands of use cases and how these foundations prepare healthcare organizations for the next evolution: agentic AI that can reason, act, and deliver real clinical value.