DagsHub

DagsHub is a web platform, built on open source tools and formats, for managing the lifecycle of machine learning and AI projects: data collection, dataset curation and annotation, experiment tracking (both model training and prompt engineering), and model management. Projects can be hosted, discovered and collaborated on in a central location, with capabilities including data versioning and lineage, notebook versioning and diffing, interactive pipelines, CI/CD/CT integration, an annotations workspace, a model registry and deployment management.

The platform is designed particularly for unstructured and multimodal data types such as text, images, audio, video, documents, medical imaging and binary files, and markets petabyte-scale data management at the enterprise tier. It integrates with users' existing storage and compute providers; data can also be hosted on the platform via DagsHub Storage, but DagsHub does not provide compute resources itself. Experiment tracking is described as compatible with MLflow, and the annotation workspace as Label Studio compatible.

Founding story

The founders started DagsHub on the premise that collaborating on data and data science problems is unnecessarily hard, because fundamental differences between data science and software development workflows make existing tools unsuitable. The stated goal was to give the ML and DS community the equivalent of the standardized project management and version control that software engineering has, making picking up someone else's work as easy as a "git checkout" and enabling open source data science.

Business model

B2B software platform sold on a per-seat subscription basis with a free tier, plus custom-quoted enterprise agreements including on-premise and VPC/air-gapped deployment.

Tiered subscriptions: a free Individual plan; a Team plan at $119 per user/month billed monthly or $99 per user/month billed yearly (14-day free trial), which adds unlimited private repositories, multimodal annotation and auto-labeling, bring-your-own storage, Label Studio compatibility, team RBAC and priority support with up to 1TB of data or 2 million files; and a custom-priced Enterprise plan adding petabyte-scale data management, deployment to the customer's cluster, full VPC/air-gapped on-premise installation, SSO/LDAP/OIDC RBAC, OpenShift compatibility, organizational resource control and an enterprise SLA.

Traction

The company states that over 65,000 data scientists use the platform, and publishes customer stories and testimonials from practitioners in applied science and machine learning roles. Its GitHub organization hosts 45 repositories, including the fds CLI (392 stars) and the DagsHub client libraries (103 stars).