Over a two-month period, contributed to both the rapidsai/cuvs and milvus-io/milvus repositories by developing features that enhance GPU-accelerated nearest neighbor search and indexing. In rapidsai/cuvs, enabled support for uint32_t as an index type in the nearest neighbor search refinement path, updating C++ templates and function signatures to facilitate a smoother transition from Raft to cuVS and improve candidate index handling. For milvus-io/milvus, configured the build system using CMake to enable GPU-accelerated knowhere-cuvs indexing and updated documentation in Markdown, clarifying GPU support and reducing configuration risks for contributors and operators. No bugs were reported or fixed.