


Durable AI runtime for teams who can't hand their data to a vendor. Training to production agents. From the team behind Flyte, the open-source orchestrator.











The full setup walkthrough: union.ai/blog-post/union-sel…

Spot VMs on GKE are cheap right up until one gets preempted mid-run. Union picks the run back up where it stopped, and now you can set it up yourself on @googlecloud ☁️ Connect your GKE cluster from the Union UI yourself. Your workflows read and write BigQuery and Cloud Storage, can call Gemini through Vertex AI, and keep their data, images, secrets and logs in your own Google Cloud project. New orgs get $950 in credits for 30 days 🎁 youtu.be/b14p8mk0nsg


We've written up how it works in two parts: Naming, versioning and lineage: union.ai/blog-post/artifacts…

Somewhere in Slack is the path to the model you're supposed to deploy. Hopefully it’s the latest one. We built Union Artifacts so datasets and models have names you can use across pipelines and apps, with versions and a record of where they came from. Publishing a version can also kick off the next pipeline with that exact version as input. Add partitions and you can process new data when it arrives or query which dates need a backfill.

The future of AI is specialized intelligence with models tuned to your data, your domain and your customers. Today, we're proud to launch live from Forge the Fireworks Frontier Ecosystem with 60+ partners across the AI stack, helping you build and ship faster. Check it out: fireworks.ai/partners