18:30 - 19:15 - AWS controllers for Kubernetes - AWS services, now Kubified

19:15 - 19:45 - Productizing ML with Kubernetes

AWS controllers for Kubernetes - AWS services, now Kubified

Do you love the Kubernetes API and user experience? Do you love declaratively defining your application as a Deployment or Daemonset, a Service, and maybe an Ingress manifest, and letting the magic of Kubernetes handle the orchestration of your application deployment? We do too! Until now, if you had a Kubernetes application with some dependencies on an AWS resources (like S3 Bucket, SNS Topic, DynamoDB Table, etc) you needed to use another tool in addition to Kubernetes, like Terraform or CloudFormation, to manage the creation and lifecycle of those resource dependencies.

Can we do better? Kubernetes API is extensible and can be extended to manage even non-kubernetes resources! In this talk we will introduce this concept, of managing cloud resources through Kubernetes API, and discuss how this concept affects current software development and deployment practices. We will also explore the AWS Controllers for Kubernetes (ACK), which lets you define your application's AWS managed service resources using the familiar Kubernetes API and manifests! No need to use a different configuration system or log into the AWS Console! Come learn about the design of the AWS Controllers for Kubernetes, what features this new project provides, and the roadmap for service integration over the coming months.

By David Feldstein

David is a software engineer with 10+ years of experience designing, building and maintaining large scale systems. Perviously David worked as a Startups Solutions Architect at AWS based in Tel-Aviv where he was responsible for creating architectural best practices and working with customers on how they use the cloud and innovation to transform their own business or disrupt new markets. Today, David is a Containers Specialists responsible for the Israeli market in AWS where he is working with customers on GTM motions and architecture best practices in regards to Containers.

Productizing ML with Kubernetes

Productizing ML is hard, and many companies are starting to build their own infrastructure for "operationalized AI" and kinda re-inventing the wheel. That being said, if you use a modern cloud-native stack, you can build your own AI Infrastructure stack fairly smoothly on Kubernetes.

In this talk, Almog, a Kubernetes maintainer, will review with us the various trends and open-source tools to build your own Production-ready AI Infrastructure stack. We'll talk about the various stages on the pipeline, about Kubeflow, KServe, Seldon, RaptorML and other ML k8s buzzwords 😎 a sneak peek into the future 👽 (which is actually... quite present).

by Almog Baku

Almog is an open-source enthusiast, he's a serial founder, a Kubernetes maintainer since 2015, and co-writer of the CI/CD Foundation's MLOps Roadmap.