In early December 2017 Google announced Kubeflow, a standard ML stack on Kubernetes, initially bundling JupyterHub and Tensorflow. This is the backstory …

I first learned about Kubeflow, a machine learning stack on Kubernetes, at the time of the KubeCon + CloudNativeCon conference in Austin, TX. Right before the conference we had an OpenShift Commons Gathering featuring an ML panel. It was at this very panel where David Aronchick (Google product manager) first publicly mentioned the launch of the Kubeflow project.


A couple of days later, David and his colleague Vish Kannan presented Hot Dogs or Not" - At Scale with Kubernetes at KubeCon + CloudNativeCon, where they motivated Kubeflow and introduced the stack:


Slide deck via Speaker Deck

The official announcement of Kubeflow followed a couple of days later via the Kubernetes blog with the post Introducing Kubeflow - A Composable, Portable, Scalable ML Stack Built for Kubernetes that the lead Kubeflow engineer at Google, Jeremy Lewi, wrote together with David.

From the blog post we also learn the great news that Kubeflow has wide support and a number of entities plan to contribute: from CaiCloud, Red Hat's OpenShift, Canonical, Weaveworks, to Container Solutions and CoreOS.

If you want to try out Kubeflow or want to get involved in its further developement, check out the following links:

I'm super excited about Kubeflow and hope you give it a spin as well!

Header photo by SpaceX / Unsplash