During December 2025, Haoyu Peng developed a batch-capable blocking retrieval path for the LMCache repository’s NixlStorageBackend, addressing the need for synchronous access to cached memory objects. He refactored locking mechanisms and optimized memory allocation, resulting in more predictable memory usage and safer concurrent access patterns. Using Python and leveraging skills in asynchronous programming, backend development, and memory management, Haoyu’s work improved the reliability of memory-object retrieval under load. The changes aligned with the platform’s caching strategy for synchronous workloads and enhanced maintainability, providing a solid foundation for future storage-backend improvements and more accurate performance characterization in concurrent environments.