DVC
Git for data — version control for datasets, models, and ML pipelines without bloating your repo
Visit DVC →Data Version Control (DVC) solves a problem Git was never designed for: tracking multi-gigabyte datasets and trained model files alongside code, without checking those large binaries directly into the repository. Instead, DVC stores lightweight pointer files in Git and pushes the actual data to remote storage like S3, GCS, or Azure.
Beyond storage, DVC defines reproducible ML pipelines as code — each stage (preprocessing, training, evaluation) is tracked with its inputs, outputs, and parameters, so a teammate can pull the repo and reproduce an exact experiment run without guessing which dataset version or hyperparameters were used.