CodaLab for Data-Driven Research


Today’s data-driven research and development is stymied by an inability for scientists and their collaborators to easily reproduce and augment one another’s experiments. CodaLab addresses this problem by providing a cloud-based virtual “workbench” where computer scientists can conduct data-driven experiments quickly and easily. These experiments can then be easily copied, re-worked, and edited by other collaborators in order to advance the state of the art in data-driven research and machine learning.

CodaLab also allows communities to create competitions to focus on some tasks, which results can then be provided back to the community as experiments for further development. By improving productivity by multiple orders of magnitude, CodaLab makes it easy for scientists and their collaborators to focus on the challenges at hand and not get delayed by the difficult and laborious work of setting up a new experiment.



CodaLab Demo Link description

CodaLab Demo


July 14, 2014


Isabelle Guyon and Evelyne Viegas


Microsoft Research, Chalearn