My own deep learning "framework". The functionality is pretty minimal, but it allowed me to start working in Rust and Web Assembly, while at the same time developing my understanding of the inner workings of how deep learning actually works.
I also decided to manually compute the derivatives required for the backward pass (instead of using automatic differentiation) in order to improve my competence with multivariate calculus.