Worked on the intel/compute-runtime repository to enhance profiling accuracy and data integrity within command queue management. Focused on low-level C++ development, addressing a critical bug in the profiling start command dispatch path. Refactored the dispatchProfilingPerfStartCommands function to ensure hardware performance counters are dispatched only when available and timestamp profiling is triggered solely when enabled. This targeted change prevents data conflicts and miscollection when multiple profiling types are active, resulting in more reliable performance analysis. Emphasized maintainability and correctness in the codebase, leveraging performance profiling expertise to stabilize profiling results and support downstream optimization efforts in complex compute environments.