is building the world's most advanced self-driving hardware and software solution. The efficiency demands of such a system require an expert model optimization (post-training quantization, layer pruning, etc) or hand-tuning GPU kernels (in OpenGL, CUDA, RocM or similar).
