I build and debug systems where failure is expensive: machine vision on live production hardware, distributed infrastructure, and the tooling that keeps both honest.
I write software for systems where being wrong is expensive: vision pipelines running on a moving production line, platform infrastructure at a $247B pension fund, services that have to stay up while traffic triples.
Most of what I'm good at is diagnostic. Working out which layer a failure actually lives in (application, container, network, or the camera itself) is usually harder than the fix, and I've gotten fast at it. The rest of the job is building the tooling so the same problem doesn't need a person the next time it shows up.
Right now that's machine vision at Tesla. Before that, distributed platform work at Ontario Teachers' Pension Plan and a few years of shipping things real users depended on. Outside of that I founded McMaster's self-driving car club, where we're building a ROS perception stack from the ground up.
Machine Vision Software Engineering Intern
Vision SystemsPalo Alto, CA
Software Engineer (Co-Op)
$247B pension fundToronto, ON
Software Developer (Contract)
E-commerce platformRemote
Founder
Software consultancyRemote
Things I built end to end. The numbers are measured, not estimated.
Concurrent Go workers claim jobs from a database queue with SKIP LOCKED so no job runs twice, with a rebuildable pgvector index and LLM checks scored against a labelled evaluation set. The full stack runs in Docker Compose with a Jenkins pipeline.
Live real-time monitoring product with paying customers, tracking 140K+ records at sub-800ms latency. Built the data pipeline, schema, backend, and Stripe billing.
Real-time distributed platform handling 1,000+ concurrent users at sub-100ms latency: events stream through Kafka into stateful services with Redis caching and horizontal scaling on Kubernetes.
Scores resume-to-job compatibility and returns structured gap analysis over a RESTful API. The inference layer runs as its own FastAPI service, decoupled from the frontend so the serving path scales independently, and end-to-end response holds under 2 seconds. Includes LaTeX editing that folds job-specific keywords back into the source document.
Grouped by what it's actually for, not by how it looks on a keyword filter.
Founder
Jan 2025 – Present
Started the club and grew it to 20+ members, building a ROS perception stack on Linux: real-time camera pipelines, publish/subscribe messaging between concurrent processes, and simulation-based edge case validation against labeled failure scenarios.
President
Sep 2024 – Present
Run a 25-person exec team across marketing, finance, and operations. 8+ events delivered for 1,000+ attendees, and a rebuilt planning process that cut event overhead 30% after tracking where the time and budget were actually going.
B.Eng. Software Engineering (Co-Op)
Hamilton, ON
Relevant coursework