Kyle Huang works on large-scale PyTorch training infrastructure at NVIDIA for end-to-end autonomous-driving models. His work includes optimizing multi-camera video loading, distributed data pipelines, and training performance on large GPU clusters using Python, C++, and CUDA.