Flexible compute

Dotscience makes self–service high performance compute directly accessible to data scientists, cutting delivery time and overheads.

Attaching any VM or local machine is as easy as point-and-shoot. Dotscience Runners exposes a familiar development environment on top of the compute: a Jupyter instance, or any IDE of choice with CLI run support.

By giving control over compute resources to data teams, Dotscience cuts model training time by orders of magnitude, and gets time-critical models into production quickly.

One-click productive ML environment

Dotscience's one-click runner setup gives data scientists the ability to access compute power when they need it. We’ll automatically provision instances in the cloud and expose a highly productive and reproducible model developmen environment with all the cutting-edge tools you need at your fingertips.

Version control for data and code

Version everything – code, datasets, hyperparameters, runtime environment – for reproducibility and easy model sharing. Our innovative Dotscience filesystem allows you to version large files and datasets at a scale impossible with existing version control systems such as GitHub.

Experiment freely as you clean datasets, add and engineer features. Collaborate via proven GitHub-like workflows.

Evidence-based model development

Dotscience can capture rich metadata from all your model runs executed in training and development.

Historical run data is made available in aggregate. Using Metric Explorer you can explore this data to get unrivaled insights into model behaviour, such as the effect of parameter choices on performance. This allows you take better decisions as you optimize and deploy your model. Collaborators working on the same model can share metadata, enabling them to learn from runs executed across the team.