Curated, ML-ready datasets from public sources → Clean CSVs, schema, metadata. Free samples + full packs.