Oloren AI ChemEngine

Robust, accurate, and reliable molecular property predictors.

Why ChemEngine?

  • Completely reproducible model architectures, data processing, training, and predictions under a common API
  • Implements leading methods to deliver the best performance
  • Native model ensembling, uncertainty quantification, and interpretability for all models

Features

Molecular Representations

Unifies molecular representations under an interchangeable API.

Interpretability

Determine relative feature importance with Shapley Value highlighting. Understand why models produce their results, transforming inference to insights.

Uncertainty Quantification

Quantify prediction accuracy with 80% confidence intervals. Uses historical results to improve future prediction accuracy.

Oloren Analysis Suite

Distributed platform for sharing OCE models, visualizations, objects, and workflows. Share with your team in a single click. Runs securely on your backend and viewable through a browser.

Benchmarks and models

Browse top-performing models on benchmarks at benchmarks.chemengine.org Integrate the best model architectures into your workflows

OCE Online

Any OCE operations can be made into a reproducible workflow. Workflows can be run on the Oloren Cloud or on your own private server (contact us for more details about private hosting at contact@oloren.ai).