Developed a feature for the ggml-org/llama.cpp repository that enables explicit configuration of the GCC version within CUDA Docker builds, addressing compatibility challenges across varied environments. This work involved customizing Dockerfiles to allow users to specify the GCC version, thereby improving the reliability and portability of CUDA image builds. By focusing on containerization and DevOps practices, the developer streamlined environment setup and reduced build failures tied to mismatched toolchains. The implementation leveraged Dockerfile syntax and CUDA toolchain management, with clear commit-based traceability. No major bugs were recorded during this period, reflecting a focused and well-scoped engineering contribution to the project.