If you want a written resource I have a blog post about the mathematics behind building a feed forward from scratch, https://max-amb.github.io/blog/the_maths_behind_the_mlp/. Kinda focuses on translation from individual components to matrix operations.