THINGS
I'VE BUILT.

This site. Built from scratch in static HTML + Tailwind with a neobrutalist aesthetic — hard ink shadows, CRT scanline overlay, pixel font labels, and a retro-modern tension that felt right for someone working at the intersection of atoms and algorithms. Includes a ternary Gibbs triangle phase diagram animation generated with Python and matplotlib.

I develop Python-based tools to support Bayesian active learning workflows for phase diagram construction, focusing on efficiently identifying zero phase fraction (ZPF) boundaries. My work involves implementing and extending multi-task Gaussian process models to jointly predict phase fractions and phase stability, while integrating physics-informed priors derived from CALPHAD simulations. I build data pipelines that connect Thermo-Calc outputs with machine learning models, design acquisition functions that target information-rich regions of thermodynamic space, and create iterative, closed-loop frameworks that can incorporate both simulated and experimental data. This approach enables faster convergence to accurate phase boundaries while reducing the number of expensive evaluations required, supporting accelerated materials discovery and optimization.