The future has landed.
And there are no hoverboards or flying cars.
Just scientists. Lots and lots of scientists.
GitHub Portfolio
A little about me
For starters, I'm a Computer Science and Math double major at the College of William & Mary, Class of 2027.
I specialize in compiler-based performance optimization for deep neural networks, high-performance C/C++ systems, and mathematical kernels.
I seek to understand the intersection of math and computer science, and how they can be applied to shape the world. Feel free to check out my work experience, research, and projects below.
Research Assistant & Software Developer
geoLab @ William & Mary
Implementing core simulation logic for Icebreaker, a strategic wargaming engine, with a team of software developers. Maintaining our containerized docker application in a collaborative Git setting.
Software Engineer Intern
@ DisinfoLab William & Mary
Worked with a team of trained software engineers to fight disinformation in the global political landscape. Co-authored a research report analyzing public sentiment trends, published in The Diplomatic Courier. Helped with initial development of Fact Forecast, a fact-checked news platform.
And these are some of my personal projects and research.
llm.cpp: the C++ LLM Inference Engine
Developed from scratch a basic C++ LLM inference engine: llm.cpp, to examine LLM forward passes under-the-hood. Implemented full transformer inference architecture (including custom tensor class and tensor operations), providing a text generation response mechanism from the pretrained model TinyLlama-1.1B-Chat-v1.0.
Directed Study in High Performance Computing
Researched various topics in high performance computing such as SIMD and SIMT instruction execution, performance optimization on CPU and GPU architectures, parallel and distributed computing, multi-threaded algorithms, memory and cache cache-efficient code optimizations in C/C++. I made use of: CUDA, MPI, and OpenMP.
Matroid Theory Implementation
`I implemented in C++ multiple forms of the matroid (an algebraic structure from combinatorial optimization) and showed empirically that matroids can solve many seemingly unrelated problems.
Youtube Judicial Comment Analysis
In Python I developed tools to gather (using the Youtube API) and clean Youtube comment data. Then I constructed a sentiment analysis tool that uses bert-base-cased huggingface transformer model. Analysis was published in The Diplomatic Courier.
Graph Theoretic Routing Engine
I developed a Waze-inspired routing engine in C++ using Dijkstra's algorithm, the A* algorithm, and a custom multithreaded A* algorithm I created. I containerized this project with docker, and deployed it using Flask API.
C++ Concurrency Study
Studied and implemented various concurrent patterns and techniques in C++. Focused on creating a performance-optimized concurrent hashmap.
"Computer Science is no more about computers than astronomy is about telescopes."
- Edsger Dijkstra