How would you optimize a slow-running data pipeline that processes large-scale genomics or clinical datasets?
