This was a full buildout for a full stack web app that integrates the Cloudinary API to act as a digital media content manager.

A relational database of more than 10k board games with searching and cross-searching methods. Automated record creation & associated record creation with a custom scraper to wrap & migrate BoardGameAtlasDB Integrated selective user authentication processes through Rails/BCypt to enable more secure and dynamic web app Implemented custom API routes to deliver multi-criterion search queries to database and serve up results Developed search methods to interact with deeply nested data structures and associated records Added email notifications through ActionMailer to deliver verification emails to new users

A data-cleansing and natural language processing research project to transform a raw dataset into a predictive model. Utilized Pandas, Seaborn to perform initial data-cleansing & visualizations to prepare for exploratory data analysis Created a feature extraction encoding layer model with HuggingFace Transformers, PyTorch to encode written essays into high-dimensional vectors for training data Built feed-forward neural network with PyTorch to be trained for classifying users into distinct user-groups (smokers/non-smokers, pet owners/non-pet owners, etc.) based on essay data alone Trained a second classifier model which operates through semantic parsing rather than feature extraction Models have slightly better predictive power than H0