Worked on the caugonnet/cccl repository, focusing on enhancing the STF module by addressing a bug in the 3-depth execution policy’s size calculation. Corrected the level index used for determining l2_size, which improved the accuracy of resource planning and reduced the risk of runtime errors in CUDA kernel execution. Developed and integrated a regression test to validate kernel configuration across multi-level specifications, ensuring ongoing robustness of the module. Utilized C++ and CUDA programming skills, along with software development and testing expertise, to deliver targeted improvements that strengthened the reliability and maintainability of the codebase during the development period.