@ferrow/ml-data-pipeline

Lightweight tabular data prep, no tensor deps: column typing, normalize/standardize with saved fit params, one-hot encoding, seeded shuffle, train/test split, simple imputation, and a chainable Pipeline that serializes its fitted params.