Learn why Two Sigma standardized on our optimization solution to scale their research efforts. Read More


Optimization Engine

Powered by the World’s Most Advanced Optimization Techniques

The core of our optimization solution lies in our optimization engine, an ensemble of global and Bayesian optimization algorithms that can effectively optimize any model, from long time-to-train deep learning models to cutting-edge trading strategies. The optimization engine intelligently chooses the best optimization technique for your problem and balances exploration and exploitation of your parameter space to obtain high-performing results within your budget.

Cutting-Edge Research

Built by optimization experts who have created some of the most widely-used open-source optimization libraries, our optimization engine combines an ensemble of different optimization techniques. Our research team is constantly working on extending the capabilities of the optimization engine by researching new optimization techniques and bridging the gap between academia and industry. To perform well across varying types of models, our optimization solution can tune mixed parameter types and supports up to 100 parameters, 10,000 distinct observations, and 100 parallel workers.

multimetric optimization

Advanced Features

Our optimization engine goes beyond traditional hyperparameter tuning packages and methods with numerous advanced features that empower modelers to accelerate model development and solve new optimization problems.

  • Multimetric Optimization facilitates the exploration of two distinct metrics simultaneously.
  • Conditional Parameters allow practitioners to define and tune architecture parameters and automate model selection.
  • High Parallelism enables you to fully leverage large-scale compute infrastructure and run optimization experiments across up to one hundred workers.

Always Improving

We constantly run quantitative empirical studies to demonstrate the excellence of our optimization engine against traditional and open-source approaches. We have built an internal evaluation framework that allows us to rigorously test, measure, and compare the impact of every single change to the optimization engine on hundreds of test and surrogate optimization problems. This framework allows us to confidently add new cutting edge optimization methods to our ensemble and ensure our solution is able to consistently achieve best-in-class performance on a wide variety, complexity, and scale of real-world problems.