Developed Docker GPU runtime support for AMD GPUs in the moby/moby repository, enabling the use of docker run --gpus with AMD hardware and extending compatibility beyond NVIDIA devices. The work involved creating a new devices_amd_linux.go implementation in Go, updating existing NVIDIA-related code, and consolidating shared runtime utilities to reduce code duplication. This approach improved maintainability and set a foundation for broader GPU ecosystem support within Docker. The project leveraged backend development skills, GPU computing expertise, and knowledge of the Linux kernel to deliver a feature that enhances Docker’s hardware flexibility for users requiring AMD GPU acceleration in containerized environments.