Worked on stabilizing GPU-accelerated test suites and runtime integration across the ROCm/jax, Intel-tensorflow/xla, and ROCm/tensorflow-upstream repositories. Focused on resolving critical bugs by addressing missing Numpy statistical function signatures and fixing runtime linking issues for RNN kernels, which previously caused test failures and unpredictable CI results. Used Python and Bazel to implement targeted bug fixes, including explicit library linking for MIOpen to ensure correct runtime behavior. Emphasized improving the reliability of statistical function tests and RNN-related workflows, resulting in reduced test flakiness and enabling smoother downstream feature development. Demonstrated strengths in bug fixing, GPU programming, and testing.