Breaking the Monolithic ML Pipeline with a Feature Store

Explore how a Feature Store for Machine Learning can revolutionize MLOps by decomposing end-to-end ML pipelines in this 35-minute talk from MLOps World: Machine Learning in Production. Learn about the separation of feature pipelines and model training/validation/deployment pipelines, their distinct requirements, preferred technologies, and management structures. Discover the benefits of implementing a Feature Store architecture, including improved efficiency and collaboration between data engineering and data science teams. Gain insights from Jim Dowling, CEO of Logical Clocks, Associate Professor at KTH Royal Institute of Technology, and lead architect of the open-source Hopsworks platform, as he shares his expertise on this innovative approach to machine learning pipelines.

Breaking the Monolithic ML Pipeline with a Feature Store