Optimal value of spark.sql.shuffle.partitions for a Spark batch Job reading from Kafka

Question

I have a Spark batch job that consumes data from a Kafka topic with 300 partitions. As part of my job, there are various transformations like group by and join which require