Developed a three-tier LRU cache system for the alibaba/PhotonLibOS repository, focusing on optimizing file caching and resource management using C++. The implementation introduced active, inactive, and idle cache tiers with adaptive thresholds, enabling dynamic management based on workload patterns. By enhancing the eviction policy to prioritize cold tiers, the work improved cache efficiency, IO throughput, and reduced memory pressure. The project also included build stability improvements for GCC 12 compatibility and refactored state tracking with string-keyed unordered sets. This performance-oriented approach provided a scalable caching foundation, demonstrating strong skills in cache management, system programming, and performance optimization.