Hands-on delivery on large-scale C++/HIP/CUDA codebases, such as ROCm (rocBLAS, hipDNN, Composable Kernel, AITemplate), the CUDA ecosystem (cuBLAS, cuDNN, CUTLASS, Thrust, CUB, NCCL), and ML framework cores such as PyTorch, TensorFlow, or JAX (C++/HIP/CUDA paths). The challenge: Help train and run AI systems that make AI itself more efficient on GPUs-tuning stacks, kernels, and workflows in ways that can materially shift what's possible on our hardware.