Worked on targeted performance enhancements for sealed-segment searches in the milvus-io/milvus repository, focusing on optimizing unfiltered query execution. Leveraged C++ and Python to streamline the search path by reducing redundant checks and utilizing bitmap data directly, which decreased latency for large sealed segments. Improved MVCC visibility handling by deriving visibility from data states, and integrated changes with Knowhere to optimize vector search paths, including hardware-accelerated bit operations. Coordinated cross-component updates to ensure traceability and prepare for future filter optimizations. Emphasized backend development and unit testing to deliver faster, more predictable search performance with minimal impact on existing users.