Your first model deployment

This page outlines the supported model journey on Aidin. Parts of the end-to-end flow are still being verified for the public launch; each step below states its status honestly.

The model journey

1. Develop and train in a notebook (or a pipeline / Katib experiment).
2. Register the model version in the model registry with its metadata.
3. Deploy the registered model to a managed inference endpoint (model serving).
4. Call the endpoint from your code, and monitor it from the dashboard.

Status by step

Step	Status
Training in notebooks	Verified
Model registry (versions, metadata, lifecycle states)	Available — registry UI verified deployed; documented behavior marked on the registry page
Registering a model from a notebook or pipeline	Pending verification — API/SDK examples will follow once the public contract is frozen
Deploying a registered model to an endpoint	Pending verification for the public journey
Calling an endpoint with authentication	Pending verification

Pending verification

The end-to-end “notebook → registry → endpoint” journey is not yet published as verified behavior. Until those pages are marked verified, treat this page as an outline, not a how-to. This documentation will be completed before the corresponding features are declared generally available — the docs will not describe behavior that has not been confirmed.

What is already confirmed

• The model registry and the KServe-based serving stack are deployed on the platform at the versions listed in Integrations.
• Endpoint types available today, and those coming soon (high-throughput LLM serving, serverless scale-to-zero), are listed on the model serving page.