Research paper: https://cs224d.stanford.edu/reports/Lewis.pdf
• Objective: This paper proposes an efficient pre-processing method for changing text data to come closer to standard English
that will improve the performance of state-of-the-art NLP models.
• Models: Vanilla Encoder-Decoder, Encoder-Decoder with Bahadanau Attention
• Implemented the Bahadanau Attention based on the research paper - https://arxiv.org/pdf/1409.0473.pdf
• Results: An average BLEU score of 0.77 with LSTM units, data augmentation, and Bahadanau Attention implementation

• Objective: To use Instakart’s outsourced data on customer orders over time to predict which previously purchased products will
be in the user’s next order.
• Models:Logistic Regression, Decision Tree Classifier, Random-Forest (ensemble), Light-GBM, XGBoost, CatBoost
• Featurization: User-related features, product-related features, User x Product-related features.
• Implemented the Custom Ensemble Classifier based on the research paper -
https://pdfs.semanticscholar.org/449e/7116d7e2cff37b4d3b1357a23953231b4709.pdf
• Results: An F1-Score of 0.365,0.371,0.367,0.377,0.376,0.379 and 0.377 in the order of models mentioned above.

• Objective: To predict whether or not a DonorsChoose.org project proposal submitted by a teacher will be approved.
• Models: Naive Bayes, Decision Trees, Gradient Boosted Decision Trees
• Featurization :For Text features - BOW,TFIDF,W2V,Sentient Analysis. For Categorical Features - One hot encoding, Response coding. For Numerical Features - Normalization.
• Results: Achieved an AUC score of 0.62,0.69for Decision Trees, and GBDT