RecessD is a startup focused on minimizing consumption of commercial buildings.
My work focuses on implementing closed loop RL minimizing energy consumption of industrial IoT systems. HVAC, boiler, and irrigation systems provide unique opportunity for savings if properly controlled.
recessd.com
Internship at Formula Solutions Inc. PLC Programming for assembly line components providing 1.8x reduction in production time for a given part. Designed 3D printed fixtures for pneumatic shears, drill press. Supervised installation of equipment on the factory floor.
Freelance work capturing photo/video for residential and commercial clients. Documenting renovations, real estate listings, and personal showcases. Creative direction in filming/editing process. Project management and outreach to clients.
Tutoring services for unviersity math/physics coursework. Operating through Queen's Eng Soc. and freelance work.
Uni Subjects: Calculus, Lin Alg., Fluid/Solid-Dynamics.
HS Subjects: Calc, Functions, Data Management, Physics.
Managed crews of 4-5 members to provide landscape-construction maintenance services during the summer.
My undergraduate thesis dissertation for Mathematics and Engineering Program. Simulated glucose-insulin interraction using non-linear dynamic modeling. Applied state-space analysis to Q-Learning in simulated environment. Awarded Keyser Prize for outstanding thesis.
Download the PDF
Developed a BCI to allow hands-free binary control of a video game via blinking. Applied signal processing and data smoothing techniques to EEG signals. Tested a variety of machine learning architectures for classification. SVM achieved CV accuracy of 94.4% for blink detection with an F1 score of 96.3%.
Download the PDF
Led team designing battery enclosures/active cooling system for electric go kart. Analyzed CAD models using flow/temperature simulations in SOLIDWORKS. Modeled and fabricated enclosures using sheet metal brake. Provided bend/weld drawings to machinists.
queensrelectric.ca
A ConvNet mapping eye movements to cursor positions allows users with a webcam to toggle page scrolling by looking at the top/bottom of the screen. We collected ~4000 training images of users looking at a mouse locations. After some preprocessing, we trained a CNN to map live video input of eyes to a location output, which can be used to trigger various actions.
Repo here