> 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.

# Overview

Models are created by training jobs.

Training uses a dataset, a workspace, a training configuration, optional model constraints, and a processing compute configuration. The result is a model that can be inspected, evaluated, and deployed.

## How model training works

Training starts from a dataset created by onboarding.

A workspace provides the context for training jobs, models, and their metadata.

A training configuration defines how induction should train the model.

Model constraints can guide which modules are included, excluded, or prioritised during training.

A training job runs the configuration against the dataset using a processing compute configuration.

The completed job creates the model used for inspection, evaluation, and deployment.

![Model training flow](https://files.buildwithfern.com/umnai.docs.buildwithfern.com/1f8937bc6c6bd3f52e03dc8e6434eee9575ec3a3033c41aa6ca4a84223264a11/assets/images/model-training-induction-flow.png)

## Training with induction

The platform currently trains models through induction.

Induction builds an explainable model from a prepared dataset. In most cases, the default training configuration is enough to start. Tune the configuration or add constraints when you need more control over model complexity, feature interactions, or training behaviour.