Yesheng Ma focused on backend performance optimization in the apache/spark repository, delivering a feature that improved the efficiency of reading large table properties. By restructuring the logic to construct properties only when numParts exists, Yesheng reduced the algorithmic complexity from O(N^2) to O(N), directly enhancing scalability and query performance for large datasets. This change led to faster property reads and reduced CPU and memory usage in SQL property handling. The work demonstrated depth in algorithmic optimization and backend development, utilizing Scala and Java to implement and review code-level changes that addressed a concrete performance bottleneck in Spark’s metadata path.