April 2026: Delivered targeted reliability and performance improvements across two core repositories, focusing on clearer error reporting, batch-operation efficiency, and memory-management correctness. In redis/redis, addressed Geo Command error handling by removing unnecessary -ERR prefixes and trailing CRLF in error formatting, improving clarity for invalid coordinates. In valkey-io/valkey, optimized HGETDEL by pausing hashtable auto-shrink during multi-field deletions to align with HDEL behavior, reducing redundant memory management work during batch operations. Also fixed a memory management bug in dismissHashtable after fork by correcting the size calculation to use sizeof(bucket) when calling zmadvise_dontneed, ensuring the full top-level bucket array is advised. These changes collectively reduce user-visible latency in batch operations, lower CPU/memory overhead, and improve safety of post-fork memory hints. Technologies/skills demonstrated include C-level hashtable internals, memory management optimization, patch review discipline, and cross-repo collaboration with explicit signed-off commits.