Hansoo Chang

As a PhD intern at Merck, I led a team of four data scientists and business analysts to develop a comprehensive data table for machine learning cluster analysis, specifically to foracast a new LDL-C drug. This project involved analyzing patient claims, financial, and biomarker data from 99 million patients using Spark SQL, R, and Python. I presented our analytics to the internal VP, where I recommended targeted marketing strategies for the drug’s upcoming launch in both the US and global markets. Additionally, I collaborated with consultants from Accenture and McKinsey & Company to present our fincial forecasts to top stakeholders.

At Delve Inc., a product design consulting firm specializing in medical and consumer products, I analyzed consumer trends using machine learning models for Samsung America for their Flip and Fold phones. This analysis was based on large-scale customer surveys, utilizing text-scraping and sentiment analysis techniques in Python. Additionally, I designed and implemented a Natural Language Processing Customer Review Analyzer for Dexcom’s G7 continuous glucose monitor, using tools like BeautifulSoup and Docker, to investigate the underlying reasons for customer satisfaction and dissatisfaction.

Data from the Pediatric HIV/AIDS Cohort Study (PHACS)

Designed a new advanced statistical framework to forecast mental health and mortality risks for children with PHIV over time in a large electronic health records (Pediatric HIV/AIDS Cohort Study). Briefly, this new algorithm incorporates the starting health condition of each child to inform future health outcomes. This method has been shown to outperform previous methods of predicting health outcomes using Biological Age.

Data from the Midlife in the US (MIDUS) Study

Developed a statistical model to incorporate both blood-based and grey matter volume to calculate both Biological and Brain Age which is used to predict multimorbidity. This is the first paper to examine both biological and brain age as a predictor for multimorbidity.

Modus Create LLC is a consulting and product development firm

I worked as a consultant with the Center for Weight, Eating and Lifestyle Science (WELL) Center at Drexel and Modus Create LLC as a PhD consultant to develop a binge-eating prediction application for smartwatches. I suggested and implemented a personalized Natural Language Processing to detect which foods are more likely to cause binge-eating episodes. I also used XGBoost to achieve above an 80% sensitivity rate to predict binge-eating episodes based on time of day, the number of calories in the food, and past episodes.

Delve Inc. (formerly Bresslergroup Inc.) is a product innovation firm specializing in consumer and medical devices

Various projects as a Data Science Intern at Delve Inc. This includes providing data visualization and summary measures for business leaders from various industries. Highlights include performing data reduction techniques and survey analysis for Samsung, usability testing for Dexcom’s CGM, and Natural Language Processing of Amazon reviews of Batiste (Church and Dwight) dry shampoo.

Designed a novel wins-loss model for Major League Baseball teams using K-Means clustering and Regression techniques. Model differed from actual results by 2.687 wins on average, beating most conventional wins-loss predictor algorithms on ESPN.

Camel Up is a board game based on horse-betting

I used Python to implement a virtual version of popular board game “Camel Up.” I then designed an algorithm to calculate the most optimal betting strategy based on game-state. Utilized to beat friends at the game.

May 2023

Invited 30 minute presentation at the annual Midlife in the US (MIDUS) Investigators Meeting at the University of Wisconsin, Madison

May 2023

Presented at Statistical Methods in Imaging (SMI) Conference at the University of Minnesota on May 2023