Comment by rxin

Did you actually read the article? It was using Spark to parallelize hyperparameter tuning, which is embarrassingly parallel.