Fine-tuning and custom models.

Custom models trained on your data, when they win.

We train a model on your examples when it beats a well-prompted off-the-shelf model on quality, speed or cost, and we show it with before-and-after tests on your own data.

What we build. Three examples of where it fits.

Domain classifier

Routes tickets, documents or leads using your categories and edge cases.

Brand-voice writer

Drafts replies and content in your tone without long prompts.

Smaller, faster model

A compact model that runs cheaper and faster on a high-volume task.

What’s included. Everything you need to launch, and keep it running.

  • Dataset preparation and cleaning
  • Baseline vs fine-tuned evaluation
  • Trained model and training report
  • Deployment to your cloud or a provider
  • Retraining plan

100% money-back guarantee. Every fixed-scope project is backed by a full money-back guarantee. The terms are written into your project agreement.

Questions. Common questions, answered.

Something else? Send us a message.

Do we need to fine-tune at all?

Often not. We start with a well-prompted model and retrieval, and only recommend fine-tuning when tests on your examples show a clear gain.

How much data do we need?

It depends on the task. For many tasks a few hundred good examples are a useful start. We assess your data before recommending anything.

Who owns the trained model?

You do. Training data, model weights where the provider allows it, and evaluation results are yours.

Ready when you are.

Tell us what you’re building. We’ll come back with ideas and next steps.

  • Free discovery call
  • Written scope and quote
  • Money-back guarantee