Falcon

Falcon is a lightweight AutoML library for Python that trains a complete machine learning pipeline from a single call. Point it at a CSV file or a pandas data frame, name the task, and it handles preprocessing, candidate model selection and evaluation, returning a trained pipeline without the usual assembly work. It is published as part of the fnnx.ai open ecosystem for machine learning. It is aimed at data scientists and engineers who need a credible baseline quickly, or who want a model they can hand to production without carrying the training environment with it. Supported tasks are tabular classification and tabular regression, with demo datasets for churn and insurance included for trying it out. Preset configurations let internal components be swapped by changing a parameter rather than rewriting code, and optional installation extras add a runtime, gradient boosting candidates and hyperparameter search. The distinguishing feature is portability: Falcon exports a finished pipeline to a single file in the FNNX packaging format or to ONNX, so the model runs on a server, in a notebook, in a browser or at the edge regardless of the frameworks used to build it. The user guide covers configuration, evaluation, export and inference. Falcon is free and open source, installed from PyPI. There is no pricing page, paid tier or hosted service on fnnx.ai. It is one of four projects in that ecosystem. The others are FNNX, an open specification and runtime that packages weights, inference code, environment and input and output schemas into one artifact; a library that puts large language models behind a familiar scikit learn interface; and a minimal deep learning framework. Each works on its own, and all of them export to the FNNX format.