Automated activity prediction models

We are developing AutoML tools for chemistry data to facilitate adoption of AI/ML

Quick baseline modeling of chemistry data can be done with LazyQSAR, our fast modeling tool. LazyQSAR produces light-weight models for binary classification and regression tasks.

Light-weight AutoML with LazyQSAR

Our flagship AutoML tool for chemistry is ZairaChem. This Python library offers robust ensemble-based modeling capabilities applicable to a wide range of modeling scenarios. At the moment, ZairaChem is focused on binary classification and regression tasks.

https://github.com/ersilia-os/ersilia-book/blob/main/book/chemistry-tools/automated-activity-prediction-models/accurate-automl-with-zairachem.md

In addition, we have developed a model distillation pipeline named Olinda aimed at producing light, interoperable models in ONNX format.

Model distillation with Olinda