Worked on the rapidsai/shared-workflows repository to restore and enhance A100 GPU CI/CD testing capabilities. Addressed the need for comprehensive test coverage by reinstating the full A100 testing matrix, ensuring compatibility across multiple Python and CUDA versions as well as various operating systems. Improved CI efficiency by enabling proxy caching for A100 GPU runners, which involved removing a conditional exclusion and aligning the workflow with the NVKS migration. Leveraged YAML and GitHub Actions to implement these changes, focusing on CI/CD pipeline reliability and speed. The work reduced risk for GPU workloads in releases by providing faster and more reliable test feedback.