During February 2026, Prateek focused on enhancing the reliability of Celeborn’s storage shutdown workflow in the apache/celeborn repository. He addressed a concurrency issue by ensuring the StorageManager’s executor service was properly shut down before awaiting termination, following Java ExecutorService best practices. This Scala-based backend fix reduced the risk of race conditions and potential data loss from concurrent RocksDB writes during shutdown, improving system stability for critical persistence operations. Prateek validated the solution against existing tests, confirming no regressions and adherence to documented shutdown semantics. His work demonstrated depth in concurrent programming and backend development using Scala and Java.