Tutorial: AI on Kubernetes Without the Chaos: Building Reproducible ML Environmen... Nourhan Mohamed

This tutorial focuses on building reproducible machine learning environments using Argo and Kubeflow on Kubernetes. Nourhan Mohamed from KodeKloud presents an open-source blueprint that addresses common challenges such as version drift and broken pipelines, which often lead to chaos in ML workflows. Participants will learn to design modular workflows for consistent reruns, track experiment lineage, and implement GitOps principles for improved auditability. The session also covers techniques like image pinning and artifact caching, as well as strategies to mitigate dependency drift and GPU scheduling conflicts, providing a comprehensive framework to enhance reliability in ML processes on Kubernetes.