Used ELMo transfer learning model from TensorFlow hub and Tensorflow
compiled from source with XLA and AVX2 support for text vector similarity
comparison for text recommendation.

Achieved good accuracy for multi-class classification by training TFIDF+ (Random Forest, XGBOOST, Naive Bayes), W2V+ (Random Forest, XGBOOST, Naive Bayes), CNN, LSTM, CNN+LSTM, Bi-LSTM+CNN on Kaggles data-set of Womens E-Commerce Clothing Reviews.

Co-authored Sentiment Classification of Indian Banks’ Customer Complaints

Trained XGBOOST to achieve 94?curacy for multi-class classification to
detect impostor on CMU keystroke data-set and communicated work to IEEE
conference.