Nicholas Hadler
I’m a chemistry PhD candidate in the Hartwig Group at the University of California, Berkeley, where I’m exploring how machine learning can accelerate discovery in synthetic chemistry.
My research brings together transition-metal catalysis, data science, and high-throughput experimentation to make catalysis faster, more predictive, and more data-driven. I’m especially interested in how domain knowledge can be used to make machine learning more effective for the natural sciences.
Selected Publications
-
Chen, J.*; Hadler, N.*; Xie, T.; Hnatyshyn, R.; Geniesse, C.; Yang, Y.; Mahoney, M. W.; Perciano, T.; Hartwig, J. F.; Maciejewski, R.; Weber, G. H. Landscaper: Understanding Loss Landscapes Through Multi-Dimensional Topological Analysis. arXiv, 2026. (*equal contribution)
-
Hadler, N.; Rinehart, N. I.; Elkin, M.; Nicolai, J.; Gheibi, G.; Chen, J.; Avaylon, M.; Maciejewski, R.; Weber, G. H.; Mahoney, M. W.; Perciano, T.; Hartwig, J. F. A 3D, Structure-Based, Deep Learning Approach for Predicting the Regioselectivity of Transition-Metal Catalysis. ChemRxiv, 2026.
Selected Projects
-
Landscaper. A Python framework for constructing, quantifying, and visualizing deep-learning loss landscapes in both low and high dimensions. With efficient sampling strategies, a suite of TDA-based metrics, and interactive visualization tools, Landscaper provides an end-to-end workflow for uncovering geometric and topological structure in modern ML models.
-
MolSelector. A lightweight web app for triaging molecular structures (.xyz, .mol, .mol2). Point it to a directory of files, inspect each molecule in an interactive 3D viewer powered by 3Dmol.js, and quickly tag it as accepted or rejected. Decisions are logged to a CSV file in the same folder for downstream analysis and version control.