A performant, modular Neural Network framework built from scratch in modern C++17 — no external ML libraries, just pure code, clean abstractions, and deep learning fundamentals implemented manually.
- ✅ Fully custom neural network with trainable layers
- ✅ Forward propagation from first principles
- ✅ Support for dense (fully connected) layers
- ✅ Activation functions (ReLU, Sigmoid, Tanh)
- ✅ Written in clean and modular C++17
- ✅ Custom matrix library with SIMD/NEON support
- ✅ Fast unit tests powered by GoogleTest
- 🧱 Designed to grow: Add backprop, optimizers, and models
This project aims to demystify neural networks by showing how they work under the hood, all without the help of TensorFlow, PyTorch, or Eigen.
Whether you're a quant, low-level dev, or just want to understand AI from first principles — this project is for you.