For most of my time at Facebook, I have been working on tools for adaptive experimentation, the machine-learning guided process of iteratively exploring a (possibly infinite) parameter space in order to identify optimal configurations in a resource-efficient manner. This work has culminated in the development of Ax – an accessible, general-purpose platform for understanding, managing, deploying, and automating adaptive experiments.

I am really proud to announce that Ax is available as an open-source library. You can check out the tool at or on GitHub.

Ax is accompanied by a sister library for Bayesian Optimization built on PyTorch, BoTorch.

I presented an overview of Ax and BoTorch at F8, Facebook’s Developer Conference. See the presentation here.

You can also read more about Ax and BoTorch on Facebook AI Blog.