Over a two-month period, contributed to ggml-org/llama.cpp and ggml-org/ggml by enhancing system robustness and memory management in C++ and CUDA environments. Developed a user interruption handling mechanism for the MTMD Vision example, introducing an interruption flag and refactoring signal handling to ensure stable, responsive inference sessions. Addressed CUDA memory allocation bugs across both repositories by correcting byte-size calculations and implementing type-safe allocation logic, which improved reliability under high inference loads. Demonstrated expertise in GPU optimization, memory management, and system programming, with a focus on cross-repository consistency and safer termination behavior during long-running computational tasks.