How to adopt cloud native machine learning with Kubernetes and Kubeflow

Description

While machine learning has caught the attention of nearly every organization looking to improve their efficiency and processes, the implementation of real-world solutions remains very challenging. Most organizations build custom-made, difficult-to-maintain platforms over many months before getting a single answer and then are required to support these fragile platforms long into the future. A year and a half ago, Microsoft introduced the Kubeflow project to make end-to-end ML pipelines on Kubernetes composable, portable, and scalable. Today, thanks to passionate contributors from all over the world, the company has the most popular ML platform for Kubernetes.

David Aronchick talks about what Microsoft has learned since building the community from scratch, how the company is able to abstract away “the hard stuff” to let developers and data scientists focus on what really matters, and where the Kubeflow project is going next. Additionally, he demonstrates how to use Kubeflow pipelines to wire together and track experiments over time, enabling advanced scenarios such as multicloud ML with continuous training and hosted services. With these tools and the support of hundreds of developers and organizations, Microsoft is able to bring ML to an entirely new audience, where as long as you can code, you can build complete end-to-end machine learning solutions.