You are tasked with optimizing the performance of a Spark batch job that processes large datasets. Using the Spark UI, you notice that the event timeline shows a high number of small tasks in a specific stage. What steps would you take to optimize this stage?