How would you identify and resolve a bottleneck in a data pipeline that relies heavily on complex data transformations?
