Flyte
the six Ws · specification
ML and data engineering teams building reliable, reproducible production pipelines at scale.
A Kubernetes native workflow orchestration platform for building versioned, reproducible data and machine learning pipelines in Python.
Self-hosted on Kubernetes, or used via Union.ai's managed hosted version.
Open sourced by Lyft in 2020, now a CNCF-hosted project actively developed in 2026.
Combines strong typing, caching and versioning with Kubernetes scale for reliable ML pipeline execution.
Requires a Kubernetes cluster plus the flytekit Python SDK to author and register workflows.
Open source under Apache-2.0, self-hosted, optional managed hosted tier from Union.ai, over 7k GitHub stars.