Bhuwan Sahni focused on enhancing the reliability of RocksDB snapshot loading in the apache/spark repository, addressing a VersionID mismatch that occurred during snapshot reloads. Using Scala and leveraging expertise in Apache Spark and backend development, Bhuwan improved the management of SST file mappings to prevent stale or incorrect mappings and reduce memory usage during reloads. The work involved removing shared state between task and maintenance threads, which increased concurrency safety and stability in production environments. By aligning SST file mappings with the latest snapshot loading workflow, Bhuwan’s contributions strengthened the robustness and maintainability of Spark’s backend storage mechanisms.