Research
Computational methods for physical systems.
My work centers on numerical simulation and computational tools for fluid flow, heat transfer, indoor environments, scientific data, and engineering decision support.
Indoor airflow and building physics
At Iowa State University, I develop scalable computational fluid dynamics workflows for K–12 classroom ventilation and indoor-air-quality analysis. The work spans simulation-ready geometry construction, meshing, incompressible flow and heat-transfer modeling, RANS/LES analysis, and passive-scalar transport.
My workflow has included an in-house C++ finite-element code, OpenFOAM, large CFD meshes, ParaView/TecPlot post-processing, and strong/weak scaling studies across Frontera, Stampede3, Nova, and cloud systems.
Simulation-ready geometry and 3D reconstruction
I work on pipelines that convert captured physical environments into computational geometry. Methods include point-cloud processing, Neural Radiance Fields, Gaussian splatting, segmentation, surface reconstruction, STL generation, and mesh preparation for CFD.
This research connects computer vision and geometry processing to a practical simulation bottleneck: constructing robust computational domains from real spaces.
High-performance scientific computing
I have conducted strong- and weak-scaling studies and large simulation campaigns using Frontera, Stampede3, Nova, LLNL’s Borax system, AWS, and Azure. My computational workflows use containerized environments and reproducible job pipelines for scientific simulation and analysis.
Selected work includes simulation outputs exceeding 50 million degrees of freedom and training at the International High Performance Computing Summer School in Kobe, Japan.
Scientific machine learning
My scientific-ML work includes Neural Radiance Fields for 3D plant geometry, reinforcement learning for classroom state inference, symbolic regression for environmental models, and domain-focused representation learning for physics simulations.
PhySiViT, developed with collaborators at Carnegie Mellon University and the Pittsburgh Supercomputing Center, was presented in the SC’25 Research Posters program.
Scientific simulation software
I also build software around scientific models. Current work includes an APSIM-centered platform for simulation configuration, experiment design, execution, qualification, provenance, and evidence-supported analysis.
The engineering goal is not to replace physics-based simulation; it is to reduce manual workflow friction while preserving reproducibility, model validity, and traceable scientific evidence.