Worked on the Mooncake repository to deliver a more robust and modular build and testing workflow for GPU-accelerated workloads. Focused on enhancing the TENT-based build system and testing configuration, the work included improvements to Redis index handling for faster data access and reduced latency. Implemented CUDA device management to optimize NVLink transport, enabling more efficient GPU communication. Conditional unit test inclusion was introduced to balance test coverage with CI efficiency, reducing build-time failures and improving feedback. Leveraged C++, CMake, and CUDA throughout the development process, laying the groundwork for scalable NVLink-enabled deployments and a more maintainable project structure.