How would you optimize a slow-running data pipeline that processes large image files for Harvard Medical School research?
