Ranked in the top 25% among 3000+ teams with a cross-entropy loss of 0.21 combining random forest with the Gated Recurrent Unit (GRU) variant of RNN for identifying duplicate question pairs, hosted on Kaggle.