Build and train genuine neural networks in PyTorch, the dominant framework in modern deep learning practice and research
You can build, train, and debug neural networks in PyTorch, understand the genuine mechanics of backpropagation and gradient descent well enough to diagnose training problems, and recognise when a deep learning approach is actually warranted over a simpler classical model.
Demonstrate genuine tensor/autograd understanding, training loop mechanics, architecture judgement, and debugging/tool-choice maturity.