DVC (Data Version Control)

DVC (Data Version Control) is an open-source version control system for machine learning projects that brings the familiar Git workflow to datasets, ML models, metrics, and experiment pipelines. Designed to work alongside Git, DVC enables data scientists and ML engineers to track changes to large files and datasets that exceed Git's practical limits, without requiring a separate infrastructure or modifying existing Git-based workflows. Files are stored in configurable remote storage backends including Amazon S3, Google Cloud Storage, Azure Blob Storage, SSH servers, and local filesystems, while Git repositories maintain lightweight pointer files that reference the actual data versions. DVC's pipeline and experiment management features allow teams to define reproducible ML workflows as directed acyclic graphs (DAGs), where each stage specifies its dependencies, commands, and outputs. This enables automatic dependency tracking, incremental recomputation, and full reproducibility of experiments. The built-in experiment tracking system records parameters, metrics, and artifacts for each run, providing comparison and visualization tools to identify the best-performing models. DVC integrates with popular ML frameworks and tools, and the companion Studio platform provides collaborative experiment tracking, model registry, and CI/CD for ML (continuous training) capabilities with visual dashboards. The project has over 14,000 GitHub stars and is used by ML teams at major organizations to manage datasets ranging from gigabytes to terabytes, track hundreds of experimental iterations, and ensure that every model can be traced back to the exact code, data, and parameters used to produce it.