Engineering design projects

This page presents a selection of my non-commercial design projects.

My design experience in industry includes the conceptual design of an electric delivery vehicle, along with the detailed design of some of its systems, such as the powertrain and the thermal management system. For another company, I also designed and analyzed train boogie structures.

Multi-speed transmission prototype 2019-21

I designed this 3-speed transmission prototype as a test bench for my PhD work. The research objective was to show that a model-based reinforcement learning algorithm can be used to iteratively learn a model of the unknown dynamics in the transmission, and tune the gearshift controller. The transmission design also became part of a patent application.

Project Beyond 2014-16

For my undergrad capstone project, I worked with a group of 21 great engineering students – both mechanical and electrical. As the technical lead, I programmed a simulator to model the energy losses during the competition, which helped answering design tradeoffs and coordinating the design efforts. We placed first at the Shell Eco Marathon Americas 2016, and were given the Best Design Award.

Baja SAE Sherbrooke 2011-13

I was part of the Baja team in Sherbrooke during my first two years of undergrad. In second year I was the design coordinator, and also designed a two-speed manual gearbox.

Vibration test bench 2016

My master's research consisted of developing a lump model for the nonlinear dynamical analysis of bolted joints in aero-engine casings. This is an experimental setup where a shaker excites a structure with a bolted joint, and we measure the response. This was used to validate the proposed lump model.

Combustion test bench 2014

This test bench was design for an undergraduate research internship, where the goal was to visualize the effect of centripetal acceleration-induced buoyancy forces on a flame. We visualized the flame front with schlieren imaging.

Data logger 2018

The goal was to design a low-cost data acquisition system to be installed on a commercial vehicle during operation, and record measurements such as GPS positions, vehicle speed, and acceleration levels at different locations on the vehicle.

Wheeled inverted pendulum in progress

This is a side project. Part of my PhD is coming up with theoretical guarantees for the safety of systems controlled with machine learning. One approach that DOES NOT provide formal guarantees but works well in practice is domain randomization. The idea is to randomly alter the physics of the simulated environment when training the controller. I thought it'd be interesting to control a wheeled inverted pendulum with a neural network as a feedback controller, and to train it on a virtual environment with domain randomization.