Projects
A concise list of projects beyond the publication record—software, engineering work, and exploratory efforts.
NovaSSM
A PyTorch reimplementation of the ShapeWorks optimizer as a sandbox for trying new loss functions and approaches to shape modeling. Autograd removes the need to write gradients by hand, and Python makes it faster to prototype ideas than extending the full ShapeWorks backend.
ShapeWorks
At the SCI Institute I contributed to ShapeWorks, an open-source suite for automated construction of compact statistical landmark-based shape models. I led implementation of shape modeling for arbitrary regions of interest—a research problem with substantial engineering depth.
Cryo-electron microscopy
Worked on automating single-particle cryo-EM grid screening with deep learning. We built a dataset with collaborators, wrote a Python/Tkinter GUI for collecting quality labels, and trained a quality-assessment network to predict labels for grid areas. See also the MICCAI 2020 paper.
EXXON CUDA GPU
Proprietary work for ExxonMobil writing CUDA code to accelerate research software for modeling geological features.
Personal & exploratory software
- CookieClicker autoplayer — screen automation with pynput; explored OpenCV template matching for real-time recognition.
- FIFAMagic — Selenium-based trading helper for FIFA Ultimate Team 18.
- Ragtime music translator — probabilistic translation of classical MIDI into ragtime-like output (Java).
- Lego VR — senior project: VR brick-building with Unity/C# and HTC Vive.
Notable class projects
- CUDA: tiled matrix multiplication and bag-of-words implementations.
- Machine learning: SVM, decision trees, and random forests.
- NLP: traditional NLP pipeline for a Kaggle challenge (strong among non-deep-learning entries).
- AI algorithms: Dijkstra, A*, minimax, and related search methods.
- Graphics: a basic ray tracer.