Kubeflow for ML workflows

Kubeflow

• Kubeflow is an open source project created by Google.
• Its an easy to use and customisable machine learning toolkit.
• Components for training models, serving models, creating workflows (pipelines).
• Anywhere you are running Kubernetes, you should be able to run Kubeflow.

Kubernetes

• An open-source container-orchestration system.
• Automate application deployment, scaling, and management.
• Can be deployed on a local machine, cloud, on-prem data center.
• Custom solutions across range of cloud providers, or bare metal environments can also be created.

Containers

• Container image is a lightweight, standalone, executable package of software
• Includes everything needed to run an application: code, runtime, system tools etc.
• Containerised application runs quickly and reliably from one computing environment to another.