Flyte is an open-source orchestrator for machine-learning pipelines, data engineering and AI agents. Workflows are written in Python as asynchronous functions, and a TaskEnvironment declares images, resources, secrets and retry policy. The runtime records the inputs and outputs of each step, so a failed run can resume instead of restarting, and for infrastructure failures such as preempted nodes it resubmits the task with adjusted resources. Flyte integrates with tools such as Spark, Ray and PyTorch. You can deploy it yourself or use an enterprise platform from Union.ai.
Flyte is licensed under Apache 2.0 and is a Graduated project of the LF AI & Data Foundation, so it is governed openly by a foundation rather than by a single company. According to the repository, the Kubernetes-native open-source backend for Flyte 2 is still to come; for a production-grade backend the project points to Union.ai. The sources consulted do not name hosting regions for Union.ai. If you self-host, your own infrastructure determines where data resides.