Orchestrating Machine Learning: Enhancing Robustness in Model Deployment

Eric Anderson (@ericmander) and Ketan Umare (@ketanumare) discuss Flyte, the open-source workflow automation platform for large-scale machine learning and data use cases. Ketan is a former engineer at Lyft, where he created Flyte to help models in Pricing, Locations, ETA, and more. Today, the project allows machine learning developers everywhere to bring their ideas from conception to production.

In this episode we discuss:

• How Flyte combines compute with parts of a workflow engine in a way that is best for the user
• The importance of reliable fares and ETA predictions at a ride-sharing app
• A progenitor to Flyte called “Better Airflow”
• Ketan’s innovative approach to bringing typing to machine learning workloads
• Why Flyte landed at the Linux Foundation

Links:

• Flyte
• Union.ai
• Apache Airflow
• Kubeflow
• Luigi
• MLTwist

Other episodes:

• Great Expectations with Abe Gong and Kyle Eaton
• Envoy Proxy with Matt Klein