I study CS at the University of Washington, with an interest in mathematics and a strong desire to learn. Here are some of my projects.
The mark, in five forms — pick one
4.3 × 10¹⁹ STATES · GRAPH SEARCH · MONTE CARLO
CS @ University of Washington
Paul G. Allen School · B.S. 2028
Seattle, WA
Work — 03
8 entries · 2021 — present
Deployed AI agent orchestration systems with MCP server plugins across internal departments serving 225+ institutional clients globally. Contributed to execution and order-routing logic in a multi-threaded C++ trading platform.
Benchmarking six context-optimization strategies — RAG, summarization, LLMLingua and hybrids — across LongBench, evaluating Llama-3.1-8B and Gemini-2.5-Pro on quality, token efficiency, latency and cost. Training an adaptive XGBoost and MLP selector to route each query to its optimal strategy over fixed baselines.
Built AppetizeAI over the weekend — an AI restaurant recommender and map interface running on AWS Bedrock foundation models.
Enhanced protein structure prediction with U-Net and diffusion-based generative models on cryo-EM data. Trained deep learning models in PyTorch on high-performance GPU clusters for molecular reconstruction and inference optimization.
Built a model predicting company profit from multi-product sales data, using XGBoost for accuracy under a tight competition window.
Coursework in data structures and algorithms, deep learning, linear algebra, probability and statistics, software design, artificial intelligence, master's-level machine learning, differential equations and computer architecture.
Imaged and deployed over 500 machines, managed inventory, and ran technical support for district staff.
Led software for an autonomous competition robot in Java and OpenCV, finishing top five in the state — and the reason I went into computer science.
Projects — 04
Ordered most technical first. Each card draws its own diagram — vary redraws it, remix adds detail.
C++ · TCP · Distributed Systems
Distributed job orchestrator with a hand-rolled thread pool and a TCP coordinator. Heartbeat health checks catch worker failures in real time, and in-progress work is rescheduled onto a live node.
C · Multithreading · HTTP
Custom linked list, hash table and query processor over a disk-based index for fast lookups across large file sets, served by a thread-pooled HTTP server. Patched XSS and directory-traversal holes.
PyTorch · TensorFlow · Angular · Flask
Generative models for protein reconstruction from cryo-EM maps — diffusion, flow matching and stacked U-Nets — with hundreds of training runs tracked and tuned in Weights & Biases.
Stack — 05
Languages
C++
Trading systems, distributed orchestration, simulation engines
Python
Deep learning research, model evaluation, data pipelines
C
Systems work — disk indexes, thread pools, HTTP servers
Java
Robotics control and coursework
Tools
AI / ML
Coursework
Certification