Sourabh Rohilla focused on improving codebase maintainability in the pytorch/torchrec repository by removing the unused memory_usage_limit_mb class variable and its associated call sites. This targeted code refactoring, implemented in Python, addressed a longstanding TODO and reduced technical debt related to memory usage handling. By eliminating obsolete code paths, Sourabh simplified future enhancements to the metric handling module and lowered the risk of confusion from stale logic. The work demonstrated a methodical approach to software maintenance, prioritizing code cleanliness and long-term sustainability. Over the month, Sourabh contributed one feature, emphasizing careful code review and alignment with project maintenance goals.