The full ML lifecycle on Kubernetes. Experiment, build pipelines, train, tune, and serve models — MLOps with Kubeflow.

Welcome to Kubeflow: Machine Learning on Kubernetes.

Kubeflow is an open-source, end-to-end machine learning platform on Kubernetes — providing the full ML lifecycle (notebooks, pipelines, distributed training, hyperparameter tuning, serving) as integrated components. This free book teaches it from the ground up: the ML platform problem and what Kubeflow is, the ML lifecycle and MLOps, Kubeflow's architecture (the integrated components), notebooks (interactive experimentation), Kubeflow Pipelines (the orchestration backbone), distributed training (the Training Operator), hyperparameter tuning (Katib), model serving (KServe), metadata/multi-tenancy/the platform (tracking, isolation, operations), and using Kubeflow in practice (adoption, the ecosystem, best practices). Ten focused chapters with clear diagrams that demystify MLOps on Kubernetes — turning ad-hoc machine learning into a disciplined, reproducible, automated, scalable end-to-end platform.

This title is part of the ShriIra library and is free to read in full, right here — our small contribution to making world-class knowledge easy to reach.

A note on reading it: open the Contents menu at the top of the reader to jump between chapters, use the Aa menu to set a comfortable text size, theme (light, sepia, or night), and single- or two-page layout. Your place is saved automatically, so you can always pick up where you left off.

We hope it serves you well.

— Shriira Press

Contents

1. Chapter 1 — What Kubeflow Is
2. Chapter 2 — The ML Lifecycle and MLOps
3. Chapter 3 — Kubeflow's Architecture
4. Chapter 4 — Notebooks: Interactive Experimentation
5. Chapter 5 — Kubeflow Pipelines
6. Chapter 6 — Distributed Training
7. Chapter 7 — Hyperparameter Tuning with Katib
8. Chapter 8 — Model Serving with KServe
9. Chapter 9 — Metadata, Multi-Tenancy, and the Platform
10. Chapter 10 — Using Kubeflow in Practice