-Cleaned and analyzed two datasets of 10,000+ patient health records in Chile to identify a strictly positive correlation between text notifications and the behavior and health of patients living with chronic diseases -Accounted for additional confounding variables (age, income, & population/hospital density) in a regression analysis, improving the MAPE metric on the percentage of people with chronic diseases in each group-Leveraged the research and data mining process to analyze… Show more -Cleaned and analyzed two datasets of 10,000+ patient health records in Chile to identify a strictly positive correlation between text notifications and the behavior and health of patients living with chronic diseases -Accounted for additional confounding variables (age, income, & population/hospital density) in a regression analysis, improving the MAPE metric on the percentage of people with chronic diseases in each group-Leveraged the research and data mining process to analyze different population characteristics between Chile and more developed countries using econometric data analysis and quasi-experimental methods Show less

-Extracted series of engineered categorical and time-based features from historical mock draft data to predict future NBA draft orderings-Analyzed methods of combining mocks that are most accurate, emphasizing the importance of accurately forecasting early draft picks by applying a log loss metric-Implemented a Naive Bayes regression classifier to predict a range of slots each player would be drafted, assigning accurate authors and subsets of authors that complement each other a higher… Show more -Extracted series of engineered categorical and time-based features from historical mock draft data to predict future NBA draft orderings-Analyzed methods of combining mocks that are most accurate, emphasizing the importance of accurately forecasting early draft picks by applying a log loss metric-Implemented a Naive Bayes regression classifier to predict a range of slots each player would be drafted, assigning accurate authors and subsets of authors that complement each other a higher weighting Show less

-Analyzed user data with A/B testing and causal analysis to build a rewards infrastructure that grew both the number of workouts and active users by over 30% by targeting the causes that led to more user activity-Improved offline RMSE metrics for the predicted number of future workouts by adjusting and weighting for outliers and over-represented users using continuous weighted regression techniques-Constructed targeted workouts for various demographics using k-nearest-neighbors… Show more -Analyzed user data with A/B testing and causal analysis to build a rewards infrastructure that grew both the number of workouts and active users by over 30% by targeting the causes that led to more user activity-Improved offline RMSE metrics for the predicted number of future workouts by adjusting and weighting for outliers and over-represented users using continuous weighted regression techniques-Constructed targeted workouts for various demographics using k-nearest-neighbors, classifying users within their region. Increased the range of user locations, classifying each new user with similar previous users Show less