- Used machine learning to make a web app that eliminates the health risks, carbon footprint, and cost of international travel by recommending local attractions that are similar to international destinations
- Leveraged NLP pattern detection over large datasets (7 million travel records and 170,000 reviews from a travel review website) scraped and cleaned in Python using selenium and beautiful soup
- Optimized recommender algorithm using topic modeling and cosine similarity in scikit-learn and nltk, improving MAP@10 by 200% and coverage by 600% compared to popularity baseline
- Engineered a front-end website with flask, HTML, and Bootstrap on AWS: http://dataispower.me