Worked on the LLNL/RAJA repository to address memory management issues specific to Windows and MSVC environments. Focused on stabilizing memory behavior by reverting a previous dynamic stack allocation fix, restoring the use of stack-allocated arrays for MSVC. This targeted change simplified the memory management path and reduced dynamic allocation overhead, aligning the MSVC implementation with standard practices used in other environments. Utilized C++ and compiler directives to ensure compatibility and maintainability across platforms. The work improved build stability and consistency for RAJA on Windows, demonstrating a careful approach to platform-specific bug resolution and memory management within a large codebase.