1. Built customer risk gradient boosting model using, Databricks, AWS, PySpark and Python that more accurately predicted the probability of a customer defaulting 2. Lead the development of model value impact analysis in an Agile environment to determine how much company revenue could be increased as a function of improved risk model performance3. Built bootstrap sampling simulation to conclude that the underperformance of customers scored using a certain risk model was random, and not the result of the model

1. Built a movie recommender system on Amazon Web Services, then used dimensionality reduction and temporal analysis to predict customer ratings2. Performed topic modeling on over 350,000 song lyrics. Afterwards clustered songs according to topic for a music recommender system3. Detected credit fraud with 93% recall and 86% precision in a highly biased (600 to 1), anonymized dataset4. Predicted the asking price of over 4,000 cars web-scraped from cars.com5. Analyzed and visualized monthly and seasonal commuter traffic at NYC subway stations

1. Confirmed with 80% confidence that certain stars were not part of the Milky Way using Monte Carlo code 2. Clustered 10,000 stars by physical metrics using K-Means Clustering3. Predicted physical metrics of galaxy simulations (10+ million rows of data) using time series analysis and K-Nearest Neighbors classifier

1. Modeled brightness of 10+ million stars in galaxy simulations using scikit-learn's Linear Regression function2. Determined best fit models using SciPy's Chi-Squared Test function and matplotlib to visualize them

1. Used Bayesian-modeling software "Pixsrc" to study to a gravitationally-lensed galaxy2. Used chi-squared testing to determine the shape of the galaxy