I work on firmware, control loops, and the integration that makes a robot run on the tenth try as well as the first. I develop mostly in C, C++ and Python, after three software internships. I finish my bachelor's at Polytechnique Montréal in May 2027. Looking for a master's project or an engineering role in robotics or embedded systems, starting summer 2027.
Detecting falls in an industrial workspace without a camera. A time of flight sensor measures depth across the zone, and the node decides on its own whether someone has fallen and stayed down. Only the alert ever leaves the device.
The whole pipeline runs on an ESP32-S3. Baseline subtraction separates people from the empty room, blob detection finds them, tracking follows them between frames, and a state machine confirms a fall before raising an alert over MQTT. The detection stages are plain C99 with no framework underneath, so they also run on a host against recorded data.
Two AgileX Limo robots explore an unknown room and build a shared map of it. The ground station only sends high level commands such as start and return to base. All navigation runs onboard. Each robot plans its own path, avoids obstacles and finds its way back to within 30 cm of its initial position.
I worked on the embedded and simulation side. The onboard control is a set of ROS2 action servers in C++. They cover mission start and stop, return to base, waypoint following, the geofence and battery management. Mapping runs on Cartographer, which fuses lidar with wheel odometry and IMU. That is what keeps drift low enough to localize reliably. Exploration uses explore_lite and navigation runs on Nav2. The two maps are merged on the robots with multirobot_map_merge. The web interface was built by other members of the team.
Mission start and autonomous exploration on the physical robots.Obstacle avoidance from onboard lidar, with no operator input.Automatic exploration planning in simulation, with the algorithm's steps shown in the interface.
libumi
Personal project, 2025
A bare-metal C library for the ATmega328P, built with avr-gcc and avr-libc. Drivers for an HD44780 LCD in 4-bit mode, USART, the internal EEPROM and matrix keypad scanning. Four worked examples and a Make based build.
eNUVIO films muscle cells contracting under a microscope. Tracer particles sprinkled on the sample make the motion visible. I was sole developer of the pipeline that tracks them across frames and turns their motion into measurements. I built the Qt desktop application the researchers use to run it and inspect the plots. Rendering runs across a process pool sized to available CPU and memory, which keeps the interface responsive. I also set up CI with static analysis, type checking and a test suite.
Python, Qt, OpenCV, NumPy.
Full Stack Developer Intern
Petal, two consecutive terms, January to August 2025
Petal builds scheduling and records software for healthcare. I worked on HUB Next-Gen, a microservice system handling care requests and patient data queries. My work spanned the Angular front end and several of the NestJS services behind it, which communicate over Kafka. Medical data follows the FHIR standard, served from an Aidbox FHIR server.
Angular, React, TypeScript, NgRx, NestJS, Ruby on Rails, Kafka, FHIR.
Software API and Process Development Intern
Ciena, May 2024 to August 2024
The internal tool's front end was built in Bonita, a business process platform with its own UI layer. I built a React proof of concept to show what a modern front end could do instead, including token based authentication.