CheMLFlow: Open-Source Platform for Cheminformatics and Materials Informatics

CheMLFlow is a new open-source tool designed to improve the workflow of scientific machine learning, particularly in cheminformatics and materials informatics. Researchers often struggle to combine data gathering, organization, representation, training, validation, screening, interpretation, and reporting into a consistent process, typically focusing on just one part. CheMLFlow simplifies this by providing modular components, ready-to-use reference pipelines, standardized results, and evaluation metrics, making it easier to benchmark different methods and datasets. It also allows for extensibility and automation with features like pluggable models, clear run artifacts, batch processing, and report generation. A comprehensive paper detailing its capabilities can be found on arXiv under the reference 2608.04942.