I've been building neural nets in Python & Pytorch for a while but felt like I truly appreciate what was happening under the covers. So I set out to build an autograd engine from scratch in Rust.

This blog is a step-by-step tutorial on how to build nanograd in about 300 lines of code. The github repo is in the blog as well.