Sai Enduri migrated the PyTorch continuous integration infrastructure from an MI300 to an MI325 GPU cluster, focusing on optimizing GPU resource allocation for testing within the pytorch/pytorch repository. Using Python and YAML, Sai preserved all existing testing plans and workflows, ensuring no test regressions during the transition. The upgrade leveraged CI/CD best practices and GPU computing expertise to enhance throughput and accelerate feedback cycles for developers. Although the work spanned a single feature over one month, the migration demonstrated careful planning and execution, maintaining stability while improving resource efficiency in the CI environment without introducing new bugs or workflow disruptions.