Introduction

Union.ai unifies your AI development on a single end-to-end platform, bringing together data, models and compute with workflows of execution on a single pane of glass.

Union.ai builds on Flyte, the open-source standard for orchestrating AI workflows. It offers all the features of Flyte while adding more capability to scale, control costs and serve models.

There are three deployment options for Union.ai: Serverless, BYOC (Bring Your Own Cloud), and Self-managed.

Flyte

Flyte provides the building blocks need for an end-to-end AI platform:

  • Reusable, immutable tasks and workflows
  • Declarative task-level resource provisioning
  • GitOps-style versioning and branching
  • Strongly-typed interfaces between tasks enabling more reliable code
  • Caching, intra-task checkpointing, and spot instance provisioning
  • Task parallelism with map tasks
  • Dynamic workflows created at runtime for process flexibility

Flyte is open source and free to use.

You can switch to the Flyte docs here.

You can try out Flyte’s technology:

For production use, you have to deploy and manage Flyte on your own cloud infrastructure.

Union.ai BYOC

Union.ai BYOC (Bring Your Own Cloud) lets you keep your data and workflow code on your infrastructure, while Union.ai takes care of the management.

Union.ai BYOC provides:

  • All the features of Flyte

  • Accelerators and GPUs (including fractional GPUs)

  • Managed Ray and Spark

  • Multi-cluster and multi-cloud

  • Single sign-on (SSO)

  • SOC-2 Type 2 compliance

Union.ai Self-managed

Union.ai Self-managed lets you keep full control of your data, code, and infrastructure.

Union.ai Self-managed provides:

  • All the features of Flyte

  • All the features of Union.ai BYOC

The only difference between Union.ai BYOC and Union.ai Self-managed is that with Self-managed you are responsible for the system infrastructure, either partially or fully, according to which option you choose:

  • Deploy and manage your data plane yourself on your infrastructure while Union.ai manages the control plane on our infrastructure.

  • Deploy and manage both your data plane and control plane on your infrastructure with support and guidance from Union.ai. This option is suitable for air-gapped deployments.