Design and implement novel neural network architectures for YOLO, run large-scale experiments and ablation studies, optimize models for accuracy and efficiency (quantization/pruning/distillation), build training and evaluation pipelines, and collaborate with R&D and engineering to productionize open-source models.