Data Science Intern
San Francisco, California, Us
- Developed a framework to train and evaluate binary classification models reducing work needed to try models on new datasets from 2 days to 30 minutes. - The coding framework performs data cleanup, feature selection, model training, evaluation, and cross-validation producing statistics, comparative visualizations, and cost-benefit analysis.- Developed in Python using SciKit-Learn, Pandas, Matplotlib, and UnitTest libraries.- Performed end-to-end analysis and hypothesis tests on user behavior and their in-game connections – presented the results with visualizations and recommendations for improving performance metrics to the product management team. - Coded the analysis pipeline in Python and SQL using a Jupyter Notebook such that it can be re-run with trivial edits. Optimized the SQL queries for speed on a columnar database.