Hello there!
I'm a computational scientist with deep expertise in machine-learned interatomic potential (MLIP) models. My work focuses on developing and deploying AI solutions to challenging problems in computational science, particularly in data engineering and highly distributed AI/ML training and inference workflows.
- cuMolFind: cumolfind - (v1.1) CUDA-accelerated Molecule Finder; reactive molecular dynamics analysis toolkit.
- ani-mm: ani-mm - OpenMM dynamics runner using ANI potential models, includes live GUI viewer and CLI utilities.
- mini-LLMs: LLMini — compact GPT-style transformer models trained on TinyShakespeare and WikiText datasets.
- Software Development: Python, CUDA-acceleration, Bash, HPC optimization, open-source contributions.
- Machine Learning: Deep neural networks, uncertainty quantification, predictive modeling and generative AI.
- Simulation: Large-scale molecular dynamics, fragment identification, reaction pathway mapping.
- Big Data Management: CuDF, Dask, and NetworkX/CuGraph for analyzing 100 TB+ trajectory datasets.
- Source Control Git/GitHub workflows, feature branching, code review.
- Visualization: VMD, PyMOL, Matplotlib, Seaborn, Plotly
-
Early Earth Hero Run Simulation
- Reactive molecular dynamics simulations on the scale of 107 atoms using ANI potentials via a LAMMPS interface.
- Scaling simulation and analysis workflows to benchmark HiPerGator's supercomputing infrastructure.
-
Graph-Based Molecular Discovery
- Detection millions of novel molecular conformations, including amino acids, dipeptides, sugars, and other prebiotic molecules.
- Expanding graph-search algorithms to identify novel compounds from large-scale, ML-driven molecular dynamics simulations
- Automating detection of reaction networks from large-scale reactive simulations.
-
LUKE: Use the Forces
- ANI model ensemble uncertainty-based conformational searching
- Select localized atomic environments from highly uncertain atomic force predictions
I am a passionate open-source contributor! Some of my projects include:
- TorchANI - Neural network potentials; accurate quantum chemical predictions at ~106 times speedup.
- LAMMPS-ANI - Implementation of TorchANI neural network potentials into LAMMPS molecular dynamics suite.
- cuMolFind - Highly scalable, CUDA-accelerated MD trajectory analysis toolkit.
- Big Early Earth Analysis - Open pipelines for large-scale molecular dynamics trajectory analysis.
- LUKE: Use the Forces - (Pre-release) Package for predictive uncertainty-based data sampling with ANI models.
-
Astronomy
- I enjoy setting up my telescope on clear nights to watch the stars, and would consider myself an amateur astronomer
- Did some undergraduate research on a binary star system in the Cygnus constellation
- Would love to create a setup to start image collection and processing at home
-
Video Games
- Who would be surprised that someone who loves simulations of physics has an interest in video games...
- Since I was a kid I have thought that the physics engines in video games are fascinating, and largely attribute my career choices to my interest in gaming.
- Mythic Depths — procedurally generated Python dungeon crawler RPG built with PyGame.
- cpp-rng-simulator — modular C++ project demonstrating RNG and sampling algorithms to implement into Mythic Depths.
