Eric Taw

Eric Taw is a 5th year PhD candidate in the Department of Chemical and Biomolecular Engineering in Jeffrey B. Neaton's group. He works on machine learning applications for novel materials discovery, specializing in metal-organic frameworks, "small"-data, and kernel methods. He obtained his BS, also in chemical engineering, from Northwestern University with honors. In his spare time, he enjoys the great outdoors, cooking, and the occasional board/video game.

One of the grand challenges of materials discovery is quantum chemical calculations that, while predictive, are too time-intensive to use for materials database screening. Ab initio calculations that make fewer approximations of the fundamental physics scale even more poorly than oft-used density-functional theory (DFT) calculations. Furthermore, current materials databases do not incorporate experimentally observed information. By incorporating computed material data from disparate sources, including the Materials Project, Eric's research will unify multiple fidelities of data (such as DFT, post-Hartree-Fock methods, and experimental data) to predict electronic properties such as bandgaps. Through incorporating all data sources and establishing cross-correlations among data and among fidelities, this method will lead to more accurate property predictions without large quantities of expensive calculations. The resulting predictions and uncertainties, which come naturally from kernel methods like Gaussian processes, will be made publicly available for all materials in the Materials Project.