I joined UBC's autonomous drone team because of my lifelong love for aviation and the magic of seeing things fly. It was the perfect way to work on systems where software has real physical consequences.

My first challenge was figuring out how to make a drone "see." I built an object detection pipeline using the DINOv3 Vision Transformer — essentially teaching the drone to spot geometric targets by comparing what it sees against a signature it's memorized, all in real-time using PyTorch.