Worked on enhancing NEON-based acceleration for deep neural network operations within the libsdl-org/opus repository, focusing on ARM64 Windows support. Updated CMake processor matching to recognize ARM architectures, enabling NEON optimizations for DNN workflows on Windows ARM64 devices. Refactored the vec_neon.h header to use the vreinterpretq_s8_s32 intrinsic, ensuring accurate data interpretation and reliable NEON intrinsic usage. These changes, implemented in C and CMake, improved performance, portability, and reliability of DNN workloads on ARM platforms. The work laid a foundation for broader NEON deployment, addressing both feature expansion and critical data handling in embedded systems environments.