I teach COMP3710, where students move from tensor and GPU fundamentals to building and training modern deep-learning models themselves.
The course covers pattern recognition foundations, convolutional architectures, segmentation, transformers, and generative models, with substantial hands-on work in PyTorch. This sits close to my own research, so I spend a lot of time helping students debug training runs, reason about what their models are actually learning, and turn a report into a defensible experimental argument.