I lead the simulation subteam in Georgia Tech's Compound Semiconductor Products VIP partnership with Sandia National Laboratories. Our modeling connects GaN strain and polarization with atomic recoil dynamics through Quantum ESPRESSO calculations and LAMMPS simulations. Static material-response calculations, controlled recoil studies, and time-step and cell-size checks establish the numerical basis for evaluating detector concepts. The scope is material response and detector feasibility, with electrical detection performance dependent on device geometry, bias, source conditions, and readout.

During my freshman year at The College of New Jersey, I worked with Dr. Michael Bloodgood on stopping criteria for active learning using small-text. The research examined when a learner can stop requesting annotations while retaining predictive performance. This experience motivated LastLabel, my subsequent independent study of published stopping methods under annotation noise and alternative stop-set constructions.

My three-year internship at Nano PV Solar provided experience in photovoltaic technology, complementing my academic interests in semiconductor research and physical modeling.

LastLabel is an independent empirical study inspired by my earlier work with Dr. Michael Bloodgood. It evaluates methods by Bloodgood and his coauthors using small-text, simulated annotation noise, and alternative stop sets. The completed study includes 15,360 trajectories, paired uncertainty estimates, budget and sampling ablations, and a technical report. Annotation savings sometimes accompany lower predictive performance, while some equal outcomes reflect shared budget fallback. Four text tasks share one corpus, and the findings are descriptive rather than a universal ranking or a new stopping algorithm. Study release and reproducible results ↗

SENSORLAB is a research concept centered on the relationship between sensor placement and thermal-property recovery. Its proposed experimental design pairs a differentiable heat model with adaptive, uniform, and random measurement layouts under equal budgets. The central questions concern parameter error, uncertainty, and identifiability from sparse observations.

LUMEN is a physically based path tracer for procedural scenes containing diffuse surfaces, ideal mirrors, and glass. Its multithreaded C++ renderer combines BVH traversal, area-light sampling, and multiple importance sampling. Deterministic sample streams and linear image output support reproducible error comparisons. The implementation also includes a CUDA wavefront backend and an OpenGL frontend for progressively displaying CPU renders. Source code and rendering benchmarks ↗