Software should feel inevitable.
I’m Bangrui — a second-year Computer Science student at McMaster University, originally from Hefei, China.
I’m drawn to problems where the difficulty isn’t writing the code — it’s understanding what’s actually happening underneath. Concurrent systems, machine learning pipelines, and the tooling around them are where I spend most of my time.
I built a handwritten digit recognizer from scratch — not to pad a portfolio, but to trace the full path from training loop to browser inference and understand where each decision matters. That’s how I approach most of my work: start from the mechanism, not the abstraction.
I care about software that is technically honest, structurally clear, and built with enough restraint to stay useful. I also photograph — mostly as an exercise in observation.
Trained a CNN from scratch, exported to ONNX, and built a real-time browser inference pipeline. No server, no API — just your device.
A deterministic simulator for a five-stage processing pipeline under three synchronization strategies — semaphores, monitors, and message passing.
Currently building systems-level tools and ML experiments. This section will grow.