How would you optimize a slow-running pandas operation when processing several gigabytes of alternative data?
