Kubeflow - recent addition in CNCF
Recently Kubeflow was accepted as CNCF incubating project. As a Kubernetes enthusiast I was curious about Kubeflow and what problem it solves.
Kubeflow is essentially a toolkit or platform developed to perform various ML Operations like data preparation, model training, model serving etc. Kubeflow leverages Kubernetes for this purpose.
In simple words, one can write various workflow tasks like data preparation, model development, evaluation and training using Kubeflow SDK as python script. When the script is executed , it generates Kubernetes manifest files in yaml format. The generated yaml, can be deployed on Kubernetes cluster.
Kubeflow dashboard provides a nice UI to visualize the ML workflow execution.
Kubeflow is cloud agnostic, all major public cloud vendor including Google Cloud, AWS and Azure provides active support by building other ecosystem components and integrating with other cloud native service like S3, Google cloud storage and BigQuery.
Kubeflow is quickly gaining popularity and will soon become graduated CNCF project.