nerd is a machine learning library built from scratch in modern C++ by a math nerd.

The goal of this project is to understand and implement the core numerical infrastructure behind modern machine learning frameworks from first principles. Rather than relying on existing tensor libraries, every major component is developed from the ground up.