Kubeflow for Machine Learning
From Lab to Production
Kubeflow is rapidly becoming the most widely used open source reference architecture for teams wanting to build and support their own AI and ML platforms. Kubeflow enables critical ML workflows such as pipelines, data preparation, and inference serving. These components are being adopted by major cloud providers in their own AI platform offerings, as well as by independent platform providers for cloud, on-prem, or hybrid installations.
The authors of this ebook—Trevor Grant, Holden Karau, Boris Lublinsky, Richard Liu, and Ilan Filonenko—were original contributors to the Kubeflow movement. In this book, they describe the fundamentals of the Kubeflow architecture in relation to its key capabilities. Their experience and insight into how these capabilities can be effectively used by anyone adopting a Kubeflow-based AI/ML platform, and is of practical value to advance the Kubeflow open source ecosystem.
One Convergence, Inc. has developed DKube, a best-in-class MLOps platform, and are proud supporters and users of the Kubeflow architecture. We are delighted to bring you three chapters of Kubeflow for Machine Learning that highlights these core components:
|Chapter 4||Kubeflow Pipelines|
|Chapter 5||Data and Feature Preparation|
|Chapter 8||Model Inference|