OCI as a Standard for ML Artifact Storage and Retrieval

Explore how Bloomberg's Data Science Platform leverages the Open Container Initiative (OCI) specification for efficient machine learning artifact storage and retrieval in this 44-minute conference talk. Discover the unique challenges of managing ML assets, including consistency, efficiency, provenance, and governance. Learn how container image registries and OCI's artifact distribution specification address these challenges through container layering, versioning, and metadata. Gain insights into Bloomberg's approach to storing and sharing ML models and datasets as OCI Artifacts, integrated throughout their platform from model building to serving. Benefit from the lessons learned and consider how to adopt similar methods or explore OCI's potential in your own ML infrastructure.

OCI as a Standard for ML Artifact Storage and Retrieval - Peyman Norouzi & Eric Koepfle, Bloomberg