If you are a machine learning and open source enthusiast, then you already know about the Kubeflow project. For the uninitiated, Kubeflow is an end-to-end machine learning platform originally developed by Google as a way to run their TensorFlow jobs that was open sourced in 2017.
The Kubeflow project is dedicated to making deployments of machine learning (ML) workflows on Kubernetes simple, portable and scalable. Our goal is not to recreate other services, but to provide a straightforward way to deploy best-of-breed open-source systems for ML to diverse infrastructures. Anywhere you are running Kubernetes, you should be able to run Kubeflow.
Since then, it has become widely adopted by both Fortune 500 companies and machine learning startups alike. It also boasts community involvement and code contributions from some of the biggest names in tech including Google, Microsoft, AWS, Intel, Red Hat, Cisco, Bloomberg, Arrikto and more.
Contribute to open source. Get paid.
Arrikto is excited to announce that we are kicking off a few initiatives this year to get more folks in the machine learning and data science community at-large, directly involved in contributing to the Kubeflow project.
The first initiative is simple. Kaggle is a popular destination for data scientists to discover Notebooks and data sets that allow them to compete for prizes while leveling up their machine learning skills in a collaborative and community-oriented environment. One of the unfortunate shortcomings of the Kubeflow project is that the vast majority of the examples in the kubeflow/examples repository are over 2 years old, require specialized components or knowledge that might prove daunting for folks just starting out with Kubeflow.
The idea here is to take a few of Kaggle’s “get started” Notebooks, whose licensing is conducive to Kubeflow’s, and see if we can make them “just work” with Kubeflow. This should make it much easier for folks who are familiar with Kaggle’s notebooks, to experience Kubeflow and run Pipeline experiments end-to-end.
So, if you are an ace Kaggle competitor, have a good working knowledge of Kubeflow or are interested in learning, Arrikto invites you to contact us to learn how you can get paid to make Kaggle Notebooks Kubeflow friendly AND contribute them back to the project in order to expand out the “examples” GitHub repo. Note that some of Arrikto’s current MiniKF tutorials are actually based on Kaggle notebooks. So, this is totally doable!
So, what’s in it for Arrikto?
At Arrikto, a fast growing and vibrant Kubeflow community is essential to our long term business success. Arrikto has been contributing code to Kubeflow since version 0.4, sits on a variety of Kubeflow development working groups, organizes 12 monthly Kubeflow Meetups and offers free weekly Kubeflow workshops and training events. We even have a dedicated developer relations team just focused on the Kubeflow community, and we are hiring!
We feel that it’s good open source citizenship to do our part to build and support the Kubeflow community. One way to do that is to build a bridge between the Kaggle and Kubeflow communities. At the same time, we also respect the fact that developers have bills to pay and deserve to be compensated for their work, open source or not. This initiative is one way to do it!
Arrikto keeps the lights on by offering Enterprise Kubeflow (EKF), an enterprise-grade distribution of Kubeflow which contains advanced capabilities for scaling deployments on EKS, AKS and GKE, plus enhanced security, data management and automation capabilities. Our hunch is that if we can make Kaggle notebooks run on Kubeflow, it won’t take much more work for our customers to experience them on MiniKF or EKF. Win-win!
Let’s do this!
Again, here’s the initiative in a nutshell:
- Are you an ace Kaggle competitor?
- Are you familiar with or interested in learning about Kubeflow?
- Do you want to make an open source contribution to the Kubeflow project?
- Do you want to get compensated for your efforts?
Interested? Then drop us a line, we’d love to chat with you!