> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.umnai.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.umnai.com/_mcp/server.

# Onboarding

Onboarding is the stage that prepares a datasource for model training.

It applies the choices that make the data model-ready: the prediction type, target, selected features, preprocessing behaviour, sharding, and encoding settings. The result is a dataset that can be used by training workflows.

## How onboarding works

Onboarding starts from a datasource created by ingestion.

The onboarding configuration defines how the datasource should be prepared for training.

The onboarding job runs the configuration against the datasource using a processing compute configuration.

The completed job creates or updates the dataset used for training.

![Dataset onboarding flow](https://files.buildwithfern.com/umnai.docs.buildwithfern.com/4d261d8ef2e787d0d3f1878c3ff7620188e17df0b40c406b82c27e3a6ab60c81/assets/images/dataset-onboarding-flow.png)

## Initial and incremental onboarding

The first onboarding job for a datasource creates a dataset.

Later onboarding jobs can update that dataset from a newer datasource. Updates point to the existing dataset `id` and can include an `increment_label` to identify the new increment.