Harnessing Data & Machine Learning

Objectives

Nearly every aspect of modern life depends on catalysts, from fuels to synthetic fibers, drugs to detergents, and paints to plastics. Current strategies for developing catalysts rely mostly on time-intensive, trial-and-error experiments. Recent advances in computer science and machine learning have the potential to speed up discovery in this field by automating search mechanisms for these vastly complex and data-rich systems, ultimately revealing hidden patterns and physical properties that scientists can use to design novel catalysts. The goal of the Leonard Lab is to develop novel data mining and extraction methodologies, which will in turn accelerate catalytic insights and innovations with potentially far-reaching advances in challenging chemistries such as water splitting, CO2 reduction, and alkane oxidation.

Projects

Large Language Model Data Extraction - . Our research utilizes large language models (LLMs) to extract and organize data from trusted, peer-reviewed publications, enabling the development of comprehensive databases for machine learning. These models are then used to identify the underlying relationships between catalyst properties, reaction conditions, and electrocatalytic selectivity, ultimately providing data-driven insights for the design of improved catalytic systems. To read more on our work with LLM data extraction read these featured articles:

Brianna R. Farris, Kevin C. Leonard; Accelerating Catalysis Understanding via Large Language Model Data Extraction and Shallow Machine Learning Techniques. JACS Au 24 November 2025; 5 (11): 5578–5589. https://doi.org/10.1021/jacsau.5c01087

Brianna R. Farris, Joshua J. Meckstroth, Kevin C. Leonard; Interpretable, low-compute machine learning integrating experimental and catalytic descriptors for sustainable CO₂ electroreduction. Green Chem. 2026; 28 (28): 11779–11789. https://doi.org/10.1039/d6gc01753c

Machine Learning for Electrochemistry - Utilizing various machine learning techniques including neural networks to deconvolute complex electrochemical reactions. These techniques focus on simulating core electrochemical techniques like SECM to solve technical challenges across different reaction mechanisms. See our most recent publication:

Darik A. Rosser, Kevin C. Leonard; High-Speed Cyclic Voltammetry Regressions Using Machine Learning. ACS Electrochem. 3 July 2025; 1 (7): 1038–1043. https://doi.org/10.1021/acselectrochem.5c00012