K8s-Native ML Workflows Meetup 2022

Anyscale and Union engineers teamed up to present Kubernetes-native workflows with Ray and Flyte. As a result of the cooperation, Flyte users can now automatically manage Ray cluster lifecycles, which includes spinning up Ray clusters on Kubernetes on-demand. Flyte also starts Ray dashboards automatically, which provides access to metrics such as memory utilization and logs across all machines.
Anyscale’s Richard Liaw
- Ray AIR, a scalable and unified toolkit for ML applications
- Reference: anyscale.com/ray-open-source
Union.ai’s Eduardo Apolinario
- Flyte and showcase the Ray integration (Slide Deck)
- Reference Article: Ray and Flyte: Distributed Computing and Orchestration