United States
Enhancing the Yelp User Experience: Explored several use cases utilizing Yelp review data, including developing a content-based recommendation engine and a text-based star generator in order to enhance end-user experience. To do so, we queried using Hive and developed a pipeline to apply various NLP methods (tokenization, Word2Vec) using Spark ML and Spark SQL. Intensive Care Unit Patient Outcome Predictions: Developed several machine learnings models (Random Forest, XGBoost, LightGBM) in Python to best predict a patient’s survival within 24 hours of admission to a hospital Intensive Care Unit based on patient history and post-admission clinical data. Data was imbalanced, so we utilized random undersampling and sample minority oversampling technique to improve the models. Drug and Clinical Trial Discovery Tool: Developed an interactive tool that connects cancer patients to potential drugs and clinical trial treatment options based on patient-specific gene mutation, cancer type, and cancer stage for several prevalent cancer diagnoses in the US. Cleaned and transformed over 1 million patient records in a custom ETL pipeline, collected using Web-APIs with Python from different sources. Developed an interactive dashboard where patients find actively recruiting clinical trials based on their individual diagnosis type and stage.