Six projects, five I designed and built, one I fought for: two autonomous robots, an offline-RL agent, two clinical ML tools, and a federal advocacy campaign.
SHT 01Robotics · ROS2 · Computer Vision
Team of 4: I owned perception + control
Indoor autonomous robot that scans, detects, drives, picks, and sorts waste across recyclable, compost, and landfill streams. Capstone project (3rd place).
SHT 02Robotics · ROS2 · Nav2 · Computer Vision
Designed & built end-to-end
A ROS2 robot that maps a home, localizes, and drives itself to a named room, planning around obstacles, flagging floor hazards with computer vision, and confirming arrival by QR code. A smart-wheelchair prototype.
SHT 03Clinical ML · Published
2nd of 7 authors, the only engineer
Peer-reviewed ML pipeline comparing 13 models on 513 patients from the SAFE-T cohort. Co-authored with Sunnybrook + University of Toronto Orthopaedics.
SHT 04Reinforcement learning
Solo course project
A Dueling DQN with a per-entity Deep Sets architecture, trained on real SportVU tracking to call shoot-or-pass, then audited until its own headline metric broke, and regrounded on real shot outcomes.
SHT 05Clinical ML
Solo project
A three-tier ML pipeline that maps CT-derived shoulder measurements onto the Walch glenoid classification, with the real trained model running live in your browser. Healthy-vs-diseased screen ~91% (AUC 0.98).
Grounded: what it’s like to be a No Fly List kid
SHT 06Advocacy · Policy
Coalition member since 2017
Long-running federal advocacy via the No Fly List Kids coalition. Toronto Star op-ed, multiple media features, direct engagement with the PMO and federal Cabinet.
In preparation · SHT 07
RideGuide, an AI transit assistant for the Greater Toronto & Hamilton Area: live arrivals, delays, and trip planning across 10 agencies, drawn from real-time GTFS-RT feeds and fronted by an LLM query pipeline. Full write-up on the way. More sheets to follow.