The full fine-tuning pipeline

Define the project.

Describe the model’s purpose, intended behavior, voice, and example format before building the training corpus.
Create training examples.

Generate examples in batches or convert source documents into training examples that match the project.
Review quality and coverage.

Inspect batches, run quality checks, and approve the examples that belong in the training dataset.
Assemble the training dataset.

Organize approved examples in the registry and prepare the training corpus and validation data.
Run supervised fine-tuning.

Choose a supported base model and run the automated training workflow on the prepared data.
Access the trained model.

When training completes, open the resulting model repository or download the available model weights.

Build a model around your use case

Use the workflow to shape a focused assistant, adapt a model to a domain, or create consistent task-specific behavior with examples you review. Start from a project definition, then move through data preparation, training, and access to the resulting model.

Create an SFT Studio Pro account

The full fine-tuning pipeline

Define the project.

Create training examples.

Review quality and coverage.

Assemble the training dataset.

Run supervised fine-tuning.

Access the trained model.

Build a model around your use case