Empowering the World’s Experts

Access the open source version of SigOpt for intelligent experimentation

Functionality

Abstract illustration bar chart

Real-world goals

SigOpt explores multiple competing metrics, including search-style experiments with many constraints.

Abstract illustration of sliding switches

Visualizations

SigOpt’s self-hosted server gives users the ability to organize, visualize and share their experiment progress.

Abstract illustration of a network of nodes

XGBoost Integration

XGBoost users can more easily and efficiently learn hyperparameters for their models using SigOpt’s integration.

Manage your own self-hosted SigOpt server for your team

SigOpt has always been designed with customer privacy in mind. With the release of our self-hosted server solution, SigOpt users can run SigOpt in whatever environment they want, and without any data leaving their servers.

 >> git clone https://github.com/sigopt/sigopt-server.git

In addition to our self-hosted SigOpt server, we also have released a version of SigOpt’s core module that can be run in-memory. Running

>> pip install 'sigopt[lite]'

lets you use the standard SigOpt connection to power a SigOpt experiment with SigOpt’s computation running locally.

Access SigOpt’s computational tooling in a lightweight installation

Quotes from Friends of SigOpt

Paul Leu

University of Pittsburgh

SigOpt has been a valued tool and SigOpt researchers have been valued partners for our additive and nanomanufacturing projects for many years. The SigOpt platform gives us access to advanced math and technology that accelerates the design of manufacturing processes.

Rafael Gomez-Bombarelli

Massachusetts Institute of Technology

Machine learning and physical sciences interplay to invent new materials, and such novel situations require tools like SigOpt to iterate quickly and surpass human intuition. I encourage everyone in my group to use SigOpt.

Marat Latypov

University of Arizona

When we came across SigOpt, we knew it would be useful in materials research. Indeed, our observations are truly expensive: be it lab experiments or physics-based simulations. SigOpt helps us with optimal designs and efficient exploration of black-box objective functions.