Worked on the apache/spark repository to improve the stability and observability of Spark’s Shuffle Service by addressing a key issue in metrics deserialization. Focused on correcting the handling of Push-Based Shuffle metrics, the work ensured accurate logging and retrieval of shuffle-related metrics within the Spark History Server. Using Scala and Spark, along with unit testing to validate changes, the fix reduced discrepancies in metric reporting across pipeline runs and maintained compatibility with existing workflows. This targeted bug fix enhanced the reliability of monitoring and troubleshooting for Spark pipelines, contributing to more dependable performance analysis and operational insight for users.