My first real project: a from-scratch Asteroids clone built in JavaScript with p5.js, extended beyond the original with a power-up system. No tutorials past the basics — I had to figure out vectors, collision, and game state on my own. It's where I learned that a "simple" game hides a surprising amount of math and architecture.

I'm an Avionics Software Engineer on UTSTAR, the University of Toronto's student team building research rockets and VTOL vehicles. I work on embedded software and multi-sensor telemetry pipelines — the systems that let a vehicle know where it is and tell the ground what's happening. It's real engineering with real constraints: hardware that fails, timing that matters, and code that can't be debugged mid-flight.

A 3D rendering engine built on p5.js's WEBGL mode, written to understand what engines like Three.js do under the hood. I implemented the transformation pipeline myself — model, view, and projection matrices, camera controls, and lighting — instead of leaning on a framework's abstractions. It permanently changed how I read graphics code.

NexaCare is a healthcare-benefits platform built in React and TypeScript that helps people actually understand and use their coverage. Benefits information is notoriously opaque — buried in PDFs and insurer jargon — and NexaCare turns it into a clear, navigable interface. It's my most complete exercise in production-grade frontend architecture.

Quantyze is a market-prediction engine that analyzes order-book structure with neural networks to forecast short-term price movement. Instead of the usual candlestick-pattern approach, it treats the raw order book — the actual standing supply and demand — as the signal. It's where my statistics major and ML interests collide with a genuinely hard, noisy problem.

NuYu is a native iOS habit and focus app built in Swift, designed around the idea that self-improvement tools should reduce friction, not add it. It combines habit tracking with focus sessions in a deliberately minimal interface. It was my deep dive into native iOS development — SwiftUI, state management, and local persistence.

Advocate is an AI legal advisor built on the Claude API, combining large language models with neural network components and benchmarked against real Supreme Court cases. The goal: make legal reasoning accessible to people who can't afford a consultation. Benchmarking against actual case outcomes kept the project honest about what AI legal analysis can and can't do.