automl-predictive-maintenance — proof of concept for machine-failure prediction on the UCI AI4I 2020 predictive-maintenance dataset (10,000 milling cycles, 3.4% failures). An AutoML leaderboard (mljar-supervised) picks the winning model family; the LightGBM model is then compiled to a dependency-free JavaScript function so the demo runs entirely in the browser on GitHub Pages — no server, no runtime.