Moksha Poladi Email & Phone Number

Cambridge, Ma, Us

• Developed a dual-score matrix strategy for client that increased conversion rate by 10% and lowered bad rate by 40%.• Automated the credit scoring process to compute credit scores for over 30 clients which decreased manual efforts by 50% (dplyr, tidyr).• Built custom underwriting models using XGBoost that leveraged traditional and alternative data that decreased bad rates by 40% on average, thus increasing the profit margin of various financial leaders and securing new clients.• Innovatively exploited a XGBoost Model to algorithmically create a set of business rules that knock-off high risk consumers and lower bad rates of remaining population by 20% on average (sklearn, pandas).

United States

• Innovatively explored methodologies and techniques used to algorithmically create business rules to predict first payment default amongst consumers.• Employed a variety of feature selection techniques to identify key attributes to be used in determining the business rules.• Used various tree-based algorithms to obtain thresholds for key attributes.• Perform data studies analyzing their loan applicants using custom TransUnion APIs and credit attributes.

United States

• Successfully completed a proof-of-concept study to show that user telemetry data can be used to screen for human behavior leading to an Intel grant for future research.• Queried 2 TBs worth of data stored on an AWS redshift cluster using a virtual machine and performed extensive EDA to observe usage patterns as a function of COVID-19. (SQL, DataGrip, Python).• Demonstrated how different user profiles behaved differently because of COVID-19 allowing Intel to perform targeted marketing on specific segments.• Conducted A/B test to show that post-COVID usage increased in a way that was statistically significant with a p-value less than 0.05.

Colombia

During my internship, I performed Exploratory Data Analysis on the company's product sales data over the past seven years. I also performed Geospatial Analysis using Kepler gl. In addition, I was responsible for creating a machine learning model to predict the sales associated with a given campaign launched by the company. This involved feature engineering using pandas and SQL.Finally, I used streamlit to create a user-friendly interface for the marketing team to use the platform to make predictions with ease.