Worked on the ai-dynamo/dynamo repository to deliver an ARM64 CUDA build configuration and establish default CUDA environment settings for the vLLM framework. This effort focused on enabling smoother deployments and improving cross-architecture compatibility, particularly for ARM64 devices. The technical approach involved customizing Docker build pipelines and overriding default settings to streamline the build process and reduce configuration friction for users. Utilizing skills in containerization, DevOps, and bash scripting, the work broadened ARM64 support and made onboarding easier for new users. No major bugs were addressed, but overall system stability improved through more predictable and default-driven configuration management.