MiniTorch is a from-scratch reimplementation of the core ideas behind PyTorch, including autodiff, tensor operations, and basic neural network modules. This project deepened my understanding of how modern deep learning frameworks work under the hood.