Worked on enhancing caching performance and reliability in the goharbor/harbor repository by refactoring the cache concurrency model. Introduced a singleflight-based approach in Go to allow concurrent requests to share a single build result, reducing contention and improving throughput for L1 cache misses. Hardened the cache path to ensure Redis L2 population even when request contexts are canceled, using advanced context and error handling techniques. Consolidated cache utilities and expanded test coverage, particularly for concurrency and error scenarios. The work demonstrated strong backend development and testing skills, delivering measurable improvements in throughput, stability, and code maintainability within a short timeframe.