talkAccelerated Machine Learning using Kubernetes and Kubeflow

Machine learning (ML) is a diverse area which requires not just data and researchers, but large amounts of flexible compute, application management and automation tools. This talk will highlight how the Kubernetes platform enables ML workloads at scale and present Kubeflow, a nascent Open Source effort, to bridge the gap between ML practitioners and infrastructure management. The possibilities and limitations of both Kubernetes and Kubeflow will be presented with a glimpse into the future. This talk will highlight some of the critical system and infrastructure design choices to be made for designing ML friendly kubernetes clusters. At the end of this talk, a ML practitioner can learn how to boost their productivity with Kubeflow and a cluster admin can learn how Kubernetes can enable ML friendly compute. Kubeflow and Kubernetes together will enable ML workload portability across private and public clouds.