Sumit Jha enhanced MongoDB Change Streams in the debezium/debezium repository by focusing on filter normalization and improving timestamp accuracy. He implemented precise regex generation in Java, normalizing whitespace in database and collection filters to ensure accurate and efficient data processing. Sumit prioritized wallTime over clusterTime for event timestamps, increasing the correctness of change event data. His work included comprehensive unit testing to validate these behaviors and addressed feedback to improve code quality and maintainability. Through these targeted backend development efforts, Sumit contributed depth in data processing, regex handling, and robust testing, resulting in more reliable and performant change stream pipelines.