Worked on stabilizing JIT instrumentation across the Intel-tensorflow/tensorflow and openxla/xla repositories by addressing unresolved symbol errors during JIT compilation. Focused on disabling coverage and profile-guided optimization instrumentation within the cc_ir_header function, ensuring reliable symbol resolution and smoother CI pipelines. Leveraged C++ and LLVM expertise to implement and propagate these fixes, using Copybara to maintain consistency between repositories. The changes reduced debugging time and improved the reliability of JIT-enabled workflows. Demonstrated a methodical approach to cross-repository collaboration, compiler design, and performance optimization, with additional contributions in Python to support build and instrumentation processes.