Size of the dataset ( 3 million samples). There were 9K+ classes,a multi-classification prob- lem.Optimised our code a lot,appied Feature Engineering and NLP techniques in python.