Workflow orchestration · Union.ai (Linux Foundation project)

Flyte

Kubernetes-native orchestrator for ML and data pipelines with strong typing between tasks, built for reproducibility at scale.

Visit Flyte Alternatives

Flyte is an open-source, Kubernetes-native orchestrator originally built at Lyft for machine learning and data pipelines, where it distinguishes itself with strongly typed interfaces between tasks: inputs and outputs are versioned and validated, which catches mismatches at compile time rather than mid-run and makes pipelines easier to cache and reproduce. Each task runs in its own container, and Flyte tracks full data lineage and versioning of every workflow execution, which is particularly valuable for ML training pipelines that need to be audited or rerun exactly. It supports dynamic workflows, map tasks for parallel fan-out, and native Python authoring via flytekit. Flyte itself is free to self-host on Kubernetes; Union.ai, founded by Flyte's original creators, offers a managed cloud version (Union Serverless/BYOC) billed on a monthly platform fee plus consumption, aimed at teams that want Flyte's guarantees without operating Kubernetes infrastructure themselves.

At a glance

Vendor Union.ai (Linux Foundation project)
Pricing model Usage-based
Free tier Yes
Deployment Cloud, Self-hosted
Open source Yes (Apache-2.0)
Best for ML and data teams needing typed, reproducible, auditable pipelines on Kubernetes.

Pricing

Flyte is free to self-host on Kubernetes; Union.ai's managed Flyte service starts at a flat monthly fee credited toward usage, with custom Enterprise pricing.

Plan Price Notes
Union.ai Team $950/month + usage $950 credited to usage monthly; 1,000 concurrent actions, $0.0075/action
Union.ai Enterprise Custom pricing 50,000+ concurrent actions, up to 1-year data retention, self-hosted control plane option

Prices read from the vendor's own page on September 21, 2026. Vendors change prices; check the source before you budget.

Features

  • Strongly typed task inputs/outputs, checked before execution
  • Kubernetes-native, each task in its own container
  • Full workflow versioning and data lineage
  • Dynamic workflows and map tasks for parallel fan-out
  • Native Python authoring via flytekit
  • Caching of task outputs to skip redundant recomputation
  • Managed cloud option via Union.ai (Team/Enterprise)

Integrations

Profile last reviewed September 21, 2026

Alternatives

Argo Workflows Kubernetes-native workflow engine that runs each pipeline step as a container, defined as Kubernetes custom resources (CRDs). Apache Airflow The de facto standard open-source workflow orchestrator, defining pipelines as Python DAGs of scheduled, dependent tasks. Temporal Durable-execution platform for building fault-tolerant application logic in code, used for microservice orchestration far beyond data pipelines.

Flyte in the index now

Terms to know

Data pipeline The automated sequence of steps that moves data from source systems to a destination, transforming it along the way. Machine learning Building models that learn patterns from data to make predictions or decisions, rather than following hand-coded rules. Data lineage A traceable record of where data came from, how it was transformed, and where it is used downstream. Reproducibility The ability of an independent researcher to obtain the same results using the original data, code and methods. Workflow orchestration Scheduling, sequencing and monitoring interdependent data tasks so each step runs only after the ones it depends on succeed.

Related guides

How to choose a workflow orchestrator Pick a workflow orchestrator by pipeline shape, who owns it, and whether you already run Kubernetes — not by the demo DAG.