No—Quilter does not train on any paying customer designs or human-created layouts. We never collect or use paying customer data to train or improve our models. Your schematic and constraints are used only to generate your board layout, and once the job is complete, that data is discarded from the training process.
Instead of learning from human examples, Quilter primarily uses a self-play approach similar to AlphaGo. The system explores many layout possibilities on its own and evaluates them with physics simulations to learn from the outcomes. This method allows us to avoid the limitations of human-designed training data and scale learning through simulation, not customer input.
This is both a core privacy promise and a technical strategy—your IP remains private, and Quilter’s performance continues to improve without relying on customer-submitted designs.
Training policy for our Non-Commercial (Free) Tier:
Keep in mind that our Terms of Use for Quilter's Non-Commercial Tier, which is free for our users, permit training on both users' direct inputs and synthetic training data generated from those inputs.
For more information on the Non-Commercial Tier Terms of Use, visit quilter.ai/terms