Leveraging Kubernetes for Machine Learning Deep Dive

With the advent of containers, Kubernetes evolved as the defacto orchestration solution to coordinate hundreds of containers at scale across a datacenter. Kubernetes opens the door for developers to access the benefits of distributed computing. As compute capacity increases relative to price we have an explosion of Machine Learning applications moving to Kubernetes. Does anybody remember “Wonder Twin powers, activate!”

You will learn how Kubernetes offers to Machine Learning an ideal orchestration tool for hosting your clever applications. We will look at common practices, containers, and deployment architectures that are common for cloud native Machine Learning. Kubeflow is one of the dominating solutions, but there are others.

  • Architectural understanding of Kubernetes from a developer perspective
  • Deploying and monitoring application on Kubernetes
  • Understanding the architecture of Kubeflow

Hands-on exercises:

  • Best practices for containerizing Python applications
  • Scaling stateful Python applications
  • Using Kubeflow
  • Leverage ML performant hardware
  • Writing, Deploying, Running and Monitoring some Machine Learning solutions

Prerequisites:

  • Just a laptop with wi-fi and a browser
  • Machine learning fundamentals
  • Basic knowledge of Python and writing containers

Jonathan Johnson

About Jonathan Johnson

Jonathan Johnson is an independent software architect with a concentration on helping others unpack the riches in the cloud native and Kubernetes ecosystems.

For 30 years Jonathan has been designing useful software to move businesses forward. His career began creating laboratory instrument software and throughout the years, his focus has been moving with industry advances benefitting from Moore’s Law. He was enticed by the advent of object-oriented design and applied it to financial software. As banking moved to the internet, enterprise applications took off and Java exploded onto the scene. Since then, he has inhabited that ecosystem. After a few years, he returned to laboratory software and leveraged Java-based state machines and enterprise services to manage the terabytes of data flowing out of DNA sequencing instruments. As a hands-on architect, he applied the advantages of microservices, containers, and Kubernetes with a laboratory management platform.

Today he enjoys sharing his experience with peers. He provides perspective on ways to modernize application architectures while adhering to the fundamentals of modularity - high cohesion and low coupling.microservices, containers, and Kubernetes to their laboratory management platform.

More About Jonathan »