this is cool !! have you experimented with getting this to run on large datasets ? IE: running this with Spark or on your datawarehouse ?