These are some of the projects that I'm currently working on or have recently worked on. My open source work can also be found on Github.
I'm currently in the process of implementing an indoor, autonomous robot. Implemented FastSLAM, A*, and DWA algorithms from first principle. Some parts of this are open source, such as
create2remix and ros-realtime-rpi4-image.
I studied the fluid dynamic interactions between a turbulent spot with a cavity with two distinct methods of numerical simulations as well as with experiments. See the paper here.
While studying fluid dynamics using experiments, I designed and built an automated sensor positioning system that achieved positioning accuracy of up to 0.025mm. This system was deployed within the wind tunnel for more than 500 hours of experiments. See Appendix A of my thesis for details (link missing due to pending publication).
As a part of my job at Shopify, I worked on Ghostferry, which is an open source library can selectively copy data from one MySQL instance to another with minimal downtime downtime. This tool has deployed into production and moved petabytes of data to balance the database shards as a part of shop mover before each year's Black Friday.
Another tool I developed while working at Shopify is mybench, which is an open source library that allows for the development of massively-parallel database benchmarks.
iTAD Omni is an omnidirectional robotics platform measuring 2 feet across capable of driving 120lb of weight. Originally developed for the iTAD fourth year project, the system now serves as a robotics research platform in Carleton University's Advanced Biomechatronics and Locomotion Laboratory.
Originally designed for Snellie, the Realtime ROS2 image builder builds a Raspberry Pi 4 image preinstalled with the
PREEMPT_RT kernel and ROS2. See the Github repository for details.
At the start of the COVID19 pandemic, I setup a camera at my house that counted the cars entering the highway. The filtering and tracking of cars are performed via a custom algorithm written with OpenCV. During the initial weeks of the lockdown, the system logged a 30% decrease in traffic. A second version of the system was put in place later in the year, and has now continuously recorded data for almost 12 months.
These are projects I've worked on in the past and have since archived.
Long ago, I worked on a news article classifier with Naive Bayes and TF-IDF.