Simply version every notebook, data file, and model in your AI project.
The Dotscience Python library allows simple annotations of notebooks & Python scripts. Unlike git based version control, the versioning happens inline, and automatically, and doesn't interrupt creative experimentation. Never lose a
For any version of any asset: code, data or model, track the provenance of that asset.
Provenance works recursively, so you can see the outputs that were fed as inputs to each pipeline stage. Provenance graph is built dynamically, no need to define the
Track parameters and summary statistics such as accuracy continuously as models are developed.
Multiple users' metrics can be shown on the same chart, allowing never-seen-before collaboration between data scientists. Unlike competitive products, metrics are tied strongly back to provenance & model, code & data
Project owners can add collaborators easily through the Settings Panel.
Collaborators see a read-only view of the project until it is forked. When forked, projects can be edited in Jupyter. When improvements to data, notebooks & metrics are made, open a Data Science Pull Requests to propose
Move AI workloads effortlessly across any infrastructure.
Start out working on your laptop, then do some testing on a GPU enabled cloud instance, and then deploy to your on-prem GPU boxes for full training. Often this would require support tickets and conversations with the ops department. With Dotscience
Get your models into production where they can start helping users.
Your models can only do so much good on your laptop, yet deploying models to production often requires lengthy interactions with other teams. Dotscience plugs straight into your CI/CD tooling, providing