Worked on the apache/auron repository to enhance the accuracy of write-time performance metrics for the CelebornPartitionWriter, a Spark extension. The approach involved introducing precise timing logic around the shuffleClient.mapperEnd operation, ensuring that the write-time metric is incremented only when mapperEnd is invoked. This adjustment improved the fidelity of performance monitoring, supporting better optimization and alignment with the BLAZE-736 initiative. The work was implemented using Scala and focused on performance tuning within the Spark ecosystem. Over the course of the month, the developer delivered this targeted feature, contributing to more reliable and actionable metrics for distributed data processing workflows.