Toronto, Ontario, Canada
In the age-old question of hardware or software, I choose both! My work bridges electrical design and data science, finding creative ways to apply software to real-world systems. As electrical lead on U of T’s unmanned aerial systems team, I dove into high-stakes power and propulsion design, hands-on manufacturing, and team leadership. I also developed a computer-aided tool that optimizes motor, propeller, and battery selection, applying research algorithms to automatically search component databases and predict flight times at different operating points. Beyond drones, I’ve built substantial software expertise through machine learning research. Last summer, I developed a Python package for Raman spectral analysis, tackling biosensing applications with real-life impact. Born and raised in Hong Kong until I was 18, and the son of two English teachers, I’ve always been drawn to people, communication, and creativity. This year I’m pursuing a minor in jazz piano, performing with the engineering jazz band, and performing as a cast member in Skule Nite (our engineering sketch comedy musical!). I’ve also been studying Mandarin since I was a kid and continue to take weekly lessons. Looking ahead, I’m eager to work at, or even found, a startup that tackles overlooked technical challenges. With my blend of hardware and software skills, international outlook, and curiosity, I’m ready to take on problems of all sizes. No challenge is too small!
Pulsenics is changing the way in which energy for 15% of the world's GDP is being used. Pulsenics develops in-line characterization and controls capabilities for the energy-intensive electrochemical industry.
Applied 15+ industry-standard data preprocessing, visualisation and data analysis algorithms to surface-enhanced Raman spectroscopy (SERS) data from our lab and from publicly-available datasets, giving key insights into collected data. Released an open-source, custom Python package called RamanLib, improving on existing packages and allowing other researchers to implement Raman analysis functions quickly and easily. Presented a podium presentation about RamanLib and Raman analysis methods at the Undergraduate Engineering Research Day (UnERD) and at biweekly lab meetings, sharing my final product with lab members and interested researchers alike and bringing attention to key developments in the field.
Surveyed current research on crystal network models, identifying an ‘equivariant’ Graph Neural Network approach for material property prediction, which enabled improved training results. Troubleshot Python coding environments, formatted datasets, and used GPU clusters for efficient model training, streamlining the computational workflow to speed up training iterations. Organized tasks using a timeline and detailed documentation, ensuring seamless knowledge transfer for future researchers.