This repository contains a full, from-scratch implementation of Neural Radiance Fields (NeRF) using PyTorch. It covers core NeRF mechanisms including positional encoding, hierarchical volume sampling, and differentiable volume rendering, trained and validated on the Lego dataset to achieve high-fidelity novel view synthesis.