In April 2026, Tasha Pais contributed to the pytorch/pytorch repository by implementing autograd support for the torch.aminmax operation, enabling both backward and forward automatic differentiation for min and max computations. Using C++ and Python, Tasha addressed a critical derivative registration bug that previously caused runtime errors, thereby improving model reliability when using torch.compile or backward passes. She updated OpInfo flags and expanded test coverage to ensure aminmax operations were fully compatible with PyTorch’s gradient-based optimization workflows. This work deepened the library’s autograd capabilities, allowing developers to leverage aminmax in advanced deep learning and machine learning models.