Date: May 5, 2022 10:00 AM Pacific Time (US and Canada)
• Welcome, Announcements & Housekeeping
• Talk #1: Intro to Kubeflow
Co-organizer, Jimmy Guerrero will give a 10-min overview of the open source Kubeflow MLOps platform. We’ll cover architecture, components, distributions and installation options.
Data scientists often need to work with lots of different libraries, languages, and applications, often with multiple versions. Conventional approaches in Kubernetes, with a legion of tailored images or a huge golden image, do not match the reality of production and prevent collaboration. In this presentation, we will see how you can leverage the concept of environment modules inside Kubernetes to make thousands of scientific libraries, languages and packages dynamically available in a simple way.
Guillaume Moutier is a Sr. Principal Data Engineering Architect at Red Hat, focusing his work on data services, AI/ML workloads and data science platforms. Former Project Manager, Architect and CTO for large organizations, he is constantly looking for and promoting new and innovative solutions, always with a focus on usability and business alignment brought by 20 years of IT management experience.
• Talk #3: Experiment Management in Kubeflow
Managing experiments is not just crucial for research teams, but also for production teams actively deploying a handful of models to thousands of models. Needless to say that experiment management (and experiment tracking) is a part of MLOps: a larger ecosystem of tools and methodologies that deals with the operationalization of machine learning. The experiments you run in Kubeflow need to be tracked for comparability and reproducibility to successfully operationalize your ML models.
In this talk, Stephen Oladele, a Developer Advocate at Neptune.ai, will introduce you to the best practices for managing your experiments in Kubeflow. You will learn how to do that natively with Kubeflow’s ML Metadata (MLMD), but also see the advantages of using third-party tracking tools like Neptune.ai.