Building things that matter, things that don't, and everything in between -- from neural networks on MRI scans to AI-controlled water gun turrets.
see my workI'm Olivier — a developer and researcher based in Ottawa, Canada. My work spans the boundary between software and science: training neural networks to detect brain tumour mutations, building automated weather-forecast pipelines, launching firmware into the stratosphere, and writing Minecraft server plugins along the way.
I studied at Lisgar Collegiate Institute, where I won Subject Awards in Exploring Technology (Grade 9) and earned an Aviation & Aerospace SHSM designation. My research on IDH mutation detection in gliomas was presented at the University of Ottawa Radiology Research Day.
A neural network that predicts IDH (Isocitrate Dehydrogenase) mutation status from brain MRI scans. Architected and implemented in PyTorch with novel architectures achieving significant improvements in detection accuracy. Findings were presented at the 34th Annual University of Ottawa Radiology Research Day.
A weather forecast trading bot for Polymarket weather markets.
An AI-powered mechanical failure detection system for aerospace applications. Moneo listens to a running propeller or engine and warns of impending failure purely from audio — no physical sensors required. Records audio at ~48,000 samples/sec, converts to spectrograms, and uses a ResNet-18 image classifier trained on healthy vs. deliberately-broken 3D-printed propellers. Achieved 100% training accuracy with excellent AUC. Extended with unsupervised anomaly detection for real-world deployment where labelled failure audio is scarce.
An AI-controlled device that detects and deters squirrels from home gardens using a water gun turret. Won four prizes at the Ottawa Regional Science Fair.
Embedded firmware for a resin curing experiment launched to the upper atmosphere as part of the Cubes in Space program, which sends student experiments aboard high-altitude balloons and rockets. Written in C++ using the PlatformIO embedded development framework.
Explaining the project to mom
Wires and whatnot
An interactive web application for exploring World War I history. Features an interactive map, a timeline interface, a data browser, and a sidebar for navigation, with user authentication. Built with React and TypeScript on Vite, backed by Firebase/Firestore for historical data.
github.com/OlivierA-Y/greatwar-interactive ↗