I primarily work on PyTorch on the architecture optimization team, my focus has thus far been expanding cuda support for quantized tensors/quantization in general.