Joseph Lee

H[AI]r Swift, SwiftUI, iOS in progress · private

A native iPhone app I'm currently building in SwiftUI. Focused on a clean, responsive user experience; more details coming as the project matures.

Chess Engine C++

Built an object-oriented chess engine in C++ as my CS246 final project, implementing full rule enforcement — castling, en passant, promotion, and check/checkmate detection — alongside a tunable AI opponent powered by minimax with alpha-beta pruning. GitHub

Quote Finder Python, Flask, React, TensorFlow

Full-stack app that surfaces relevant quotes and verses from natural-language prompts using sentence-transformer embeddings and cosine similarity, served through a Flask REST API and consumed by a React front-end. Reached 85% relevance accuracy on a held-out evaluation set. GitHub

Image Lab Python, Tkinter, Flask, OpenCV, Docker

Desktop image-processing app exposing 10+ OpenCV transformations through a Tkinter GUI, backed by a containerized Flask REST API and verified with Postman. Profiled and rewrote the processing pipeline to bring end-to-end runtime from 2.5s down to under 1s, and integrated Real-ESRGAN to clean up low-quality uploads. GitHub

Magestic Brawl Python, Pygame-CE Gymnasium currently adding new features and reinforcement learning

A 2-d streetfighter game that uses magestic characters. Written in Python and I try to improve it whenever I have the time. GitHub

ML Notes Markdown, Python (Pandas, Tensorflow, Numpy, PyTorch)

An evolving ML/AI learning sandbox with study notes, notebook experiments, and chapter-by-chapter work from hands-on machine learning resources. GitHub

Notes Markdown

An open, growing collection of notes spanning coursework, papers, tools, and ideas I want to keep close at hand. Public so others can read along or borrow from it. GitHub

ECCC Scripts Python

During my co-op at Environment and Climate Change Canada, I made a few scripts to help boost my productivity at work. I changed the scripts to be generic since I can't display any data that I used while working. GitHub

Self-Driving Race Car Python · PyTorch · Gymnasium

Rebuilding a deep reinforcement learning agent that learns to drive a race car from raw pixels. The agent uses DQN to navigate a procedurally generated track in the CarRacing-v3 environment, with the goal of scoring 900+ averaged over 100 races. A rewrite of an existing TF1 implementation, now in modern PyTorch. GitHub

Neural Ronin Python · Pygame · Gymnasium · Stable-Baselines3 in progress

A 2D samurai platformer built as a reinforcement learning training environment. The game is built from scratch using the pygpen engine, featuring sprite animations, physics, tilemap-based levels, and combat. A Gymnasium wrapper exposes the game as an RL environment, with a PPO agent trained via Stable-Baselines3 to learn to fight through levels from game state observations. GitHub

Study Room Python · FastAPI · WebSockets · SQLite · Vanilla JS in progress

A real-time co-working web app where people create a room, share the link, and get a synced Pomodoro timer, live presence, a sticky-note corkboard, and Spotify "now playing" cards — no accounts, the unguessable room id in the URL is the only key. Built on FastAPI with WebSocket broadcasts for per-room state, the timer stays in sync without per-second server ticks by broadcasting only the end time and letting each browser correct for clock skew locally. Persistent data lives in SQLite while ephemeral presence stays in process memory. Deployed on Fly.io. GitHub