Physics-Guided 3D Gaussian Splatting
Framework coupling differentiable fluid simulation with 3D Gaussian representations for physically functional object generation.
Read the PG-3DGS paper on arXiv
3DGS / Differentiable Simulation / PyTorch / CUDA
Personal Profile
I am a PhD student in Computer Science at Purdue University working on 3D generative models for physically functional objects.
I design models that generate and reconstruct 3D objects with plausible appearance and useful physical behavior. My work focuses on learned 3D representations that account for geometry, materials, and simulation, with the goal of producing objects whose structure supports their intended function.
Recent Paper
PG-3DGS: Optimizing 3D Gaussian Splatting to Satisfy Physics Objectives
PG-3DGS connects appearance-based 3D reconstruction with physics-based reasoning. The framework adds differentiable physics objectives to Gaussian Splatting, guiding generated shapes toward physical functionality while preserving visual quality.
We physically validate our method by 3D printing generated planes optimized for lift, and measuring their lift under identical airflow conditions.
| Submitted | May 11, 2026 |
|---|---|
| Comments | Submitted to Artificial Intelligence Journal. 52 pages. |
Selected Work
Missing Soil Property Imputation
Colorado School of Mines MS thesis work using machine learning to impute missing soil properties for engineering datasets.
Machine Learning / Data Imputation / Numerical Data
Orebody Data Modeling
MInDS Lab research implementing and improving deep learning methods, including Generative Adversarial Imputation Networks, for missing orebody data.
GANs / OOD Learning / Multi-Instance Learning
Product Recommendation System
Field session project with Venvee: built a PyTorch model and AWS-backed API to predict purchase categories from user data.
PyTorch / Pandas / AWS / Training Pipelines
Education, Research, and Experience
| When | Where | What |
|---|---|---|
| 2024-current | Purdue University | PhD in Computer Science. Research in AI, 3D Gaussian splatting, differentiable simulation, and physics-based modeling. |
| 2022-2023 | Colorado School of Mines | MS in Computer Science, 4.0 GPA. Thesis: Imputing Missing Soil Properties Using Machine Learning. |
| 2019-2022 | Colorado School of Mines | BS in Computer Science, Data Science focus, Computational and Applied Mathematics minor, 3.9 GPA. |
Research Experience
- Purdue AI research on physically functional 3D objects from images.
- MInDS Lab deep learning methods for missing orebody data.
- Mining Engineering Department models for mine safety feature analysis.
Industry Experience
- Payne Institute Earth Observation Group satellite-data preprocessing and log analysis.
- nou Systems backend and simulation scheduler work for the Missile Defense Agency.
- Python, SQL, GraphQL, Flask, PostgreSQL, Linux, and web scraping.
2024 Presidential Excellence PhD Award, Purdue University