DVC: Git for data and models
Git stores code. Data—gigabytes of CSV, images, model weights—does not belong in Git. DVC replaces a data file with a tiny `.dvc` pointer committed to Git while the blob lives in remote storage (S3, GCS, Azure, SSH). Git checkout plus `dvc pull` restores the exact bytes. Reproducibility is end-to-end: code, data, and pipeline are pinned in one commit.