Worked on stabilizing the speculation testing pipeline for the EPFL-LAP/dynamatic repository, focusing on improving the reliability of integration tests. Addressed issues arising from recent MLIR syntax changes and corrected timing model path discrepancies to ensure accurate test execution. Updated operator alignment in the subdiag and subdiag_fast tests by replacing the load5 operator with load2, which reduced test failures and improved suite stability. Utilized skills in bug fixing, integration testing, and scripting with MLIR and Python to enhance the correctness of speculation-driven optimizations. The work resulted in fewer flaky tests and increased confidence in the testing infrastructure’s robustness.