The Motivation for MLOps - Architectural Perspective

Explore the motivation behind MLOps in this 57-minute talk by Steven Fines, Sr. Principal ML Architect at CoreLogic, hosted by Ben Epstein at MLOps Community Meetup #118. Delve into key areas driving MLOps adoption in enterprise ML solution development and operations. Learn about compliance challenges, model management, operational concerns, and the architectural viewpoint of MLOps. Discover the relationship between MLOps and DevOps, prerequisites for implementation, foundational components, and scaling processes. Gain insights on standardization, moving model execution to pipelines, result monitoring, and when to create data catalogs or feature stores. Understand classic obstacles in ML that differ from traditional software development, and benefit from Fines' 26 years of software engineering experience, with 15 years focused on ML and analytics pipelines.

[] Introduction to Steven Fines
[] Highlight
[] Motivating to MLOps
[] Machine Learning in the wild
[] Compliance challenges
[] Model Management
[] Operational concerns
[] MLOps: Definition from the Architectural viewpoint
[] MLOps and DevOps
[] Do I need to do this?
[] What is the framework?
[] Prerequisites
[] Foundational Components
[] Scaling the process
[] Fully mature
[] Implementing MLOps
[] Standardize
[] Move model execution to pipelines
[] Monitor the results
[] Scales Steven worked with before
[] When to create a data catalog or a feature store
[] Data catalogueing
[] Classic obstacles in ML that doesn't present in classical or traditional software
[] Wrap up