Kartik Sirohi contributed to the pytorch/pytorch repository by addressing a critical bug in the torch.clamp function, focusing on the correct handling of NaN values as scalar bounds. Using both C++ and Python, Kartik implemented a fix ensuring that when a NaN is provided as a bound, the output propagates the NaN rather than returning the input unchanged. He reinforced this behavior by adding a dedicated regression test, emphasizing robust unit testing and careful attention to numerical edge cases. This work improved the reliability of numerical computing in PyTorch and reduced the risk of silent errors in downstream pipelines.