LlamaNet

Built a neural network deep learning library in C++ entirely from scratch, using only the standard library with no external dependencies. It supports fully customizable network architectures, including configurable layers and activation functions like ReLU, Sigmoid, and Tanh, along with Adam optimization, Xavier initialization, and mini-batch gradient descent with backpropagation.

I designed the framework to be modular and scalable, making it easy to adapt to different topologies and datasets. Tested on the MNIST handwritten digit classification benchmark, it reached over 98% accuracy across 10,000 test cases.