As organizations adopt Large Language Models (LLMs) and AI-driven workflows, managing model versions, lineage, and governance across hybrid and multi-cloud environments becomes complex. Existing model registries are often vendor-tied or lack interoperability with cloud-native CI/CD and GitOps systems.
In this session, we introduce ModelPack, a core component of the CNCF-incubated KitOps project, that standardizes model packaging, metadata, and promotion workflows — enabling a vendor-neutral enterprise model registry.
We’ll demonstrate how KitOps brings GitOps-style principles to MLOps — from version-controlled model promotion to secure artifact delivery — seamlessly integrating with Kubernetes, OCI registries, and open metadata standards.
Attendees will learn how to design reproducible, auditable, and compliant AI pipelines using open, declarative ModelOps practices.