Liner Developer Platform

Liner is a free, locally-run desktop tool that takes labeled training data and produces an exported ML model without requiring any code or ML expertise. The workflow is three steps: import your data or pull from the built-in dataset library, press train and let Liner select and run an appropriate model automatically, then export the result to a platform your application can consume. Supported task types include image classification, text classification, audio classification, video classification, object detection, image segmentation, and pose classification — all selectable from project templates at the start of a session.

The defining characteristic of Liner is that training runs entirely on the local machine, including on standard CPUs without a dedicated GPU. The vendor states models are optimized for CPU training and that the data never leaves the computer. For teams handling sensitive documents, proprietary audio, or regulated imagery, this means the model development cycle stays fully air-gapped from any cloud service — a constraint that cloud-hosted AutoML tools cannot satisfy by design.

Liner fits a narrow but real use case: a domain expert who has labeled data and needs to test whether that data can support a classifier, without a machine learning engineer in the loop. It also fits privacy-sensitive prototyping where the cost of any cloud data transfer is unacceptable. It breaks down when a team needs to iterate on model architecture, tune hyperparameters, version experiments, integrate training into a CI/CD pipeline, or serve predictions to an external API — none of which the tool surfaces. Teams that outgrow the prototype stage will find no migration path inside Liner and will rebuild using a framework like PyTorch, TensorFlow, or a hosted AutoML service.

Exported models are described as edge-optimized and compatible with mobile and edge deployment targets, though the vendor page does not specify which export formats or runtimes are supported. The application is distributed as a native download for macOS (Intel), macOS (Apple Silicon), and Windows, and the vendor states it is free with no feature gating mentioned.