Worked on enhancing audio processing robustness and backend compatibility across C++ projects. In Mintplex-Labs/whisper.cpp, refactored audio buffer length handling by centralizing the calculation of buffer positions and lengths, improving reliability when buffers wrap and reducing potential runtime errors. In ggml-org/llama.cpp, addressed quantization workflow issues by implementing an automated CUDA-to-CPU fallback for clip quantization and updating model loading calls to maintain compatibility. Focused on error handling, buffer management, and GPU programming, these changes reduced support overhead and improved deployment reliability. The work demonstrated a methodical approach to modernizing core components and ensuring smoother developer and user experiences.