May 2022 – Jan 2024 · 1 yr 8 mos · Chicago, Illinois, United States · Hybrid

• I was recruited to the Fortune 16 Healthcare organization to contribute towards statistical machine learning, propensity modeling, statistical modeling, data wrangling and reporting to drive marketing strategies for Walgreens Boots Alliance. I worked on cross-portfolio and retail insights projects and I was part of the Customer Science team at Walgreens Boots Alliance.
• TECH STACK - PySpark, Python, R, SQL, PowerBI, Microsoft Azure Databricks
• SELECTED ACHIEVEMENTS:
• ✔ Built and deployed a successful Credit Card Propensity Tool prediction framework (propensity model) using XGBOOST (multi-class classification) for Walgreens Front of Store Customers in Python. The model had high accuracy, precision, and recall, and it generated an increase of 35% in Walgreens Credit Card transactions.
• ✔ Built a Customer Behavioral Progression propensity model for Walgreens Front of Store in Python using XGBOOST for multi-class classification. Used insights from the last decision tree in the XGBOOST ensemble to make recommendations to the business. This was a highly ambiguous project with changing requirements. Explored classification, regression, and constrained optimization based approaches to align with changing business requirements.
• ✔ Wrote complex and efficient SQL queries processing > 40B records to successfully complete ad hoc requests and fire drill requests from the business, within short timeframes.
• ✔ Performed K-means clustering in PySpark to efficiently segment a dataset of customers into groups based on customer behavior metrics.
• ✔ Reviewed SQL, Python and Pyspark code written by junior associates and gave them best practices recommendations.
• ✔ Mentored and guided junior associates. Managed external consultants.
• ✔ Acted as the lead interviewer for Associate Data Scientist roles in the company.

Oct 2021 – Apr 2022 · 6 mos · Chicago, Illinois, United States

• I was recruited to the organization to contribute towards machine learning, A/B testing, customer segmentation analytics, statistical modeling, data wrangling and reporting to drive marketing strategies for Holland America Line.
• SELECTED ACHIEVEMENTS:
• ✔ Creating interactive dashboards and reports using Tableau.
• ✔ Automating Excel reports in R.
• ✔ Applying the k-means() algorithm in R for optimal customer segmentation.
• ✔ Data wrangling in R using dplyr, data.table, tidyr, tidyverse, and lubridate packages.
• ✔ Write complex SQL queries fetching data with >8M rows using TOAD for Oracle.
• ✔ Manipulating data using Pivot Tables in Microsoft Excel.

Mar 2021 – Aug 2021 · 5 mos · Chicago, Illinois, United States

• I was recruited to the organization as a Data Scientist at Amwins Specialty Casualty Solutions (ASCS).
• SELECTED ACHIEVEMENTS:
• ✔ My day-to-day is fitting statistical distributions to insurance claims data, and running goodness of fit tests on the data. R packages used – dplyr, data.table, fitdistrplus
• ✔ Performed principal component analysis (PCA) on a matrix of numeric tf_idf vectors calculated from raw text data, reducing 3214 columns to 100 principal components. R packages used - tidytext, dplyr, data.table, prcomp (function in stats package)
• ✔ Built a XGBOOST multi-class classification model on top of insurance claims data after PCA and achieved 91% accuracy on test data after 10-fold cross validation and using text mining techniques in R. R packages used - tidytext, dplyr, data.table, xgboost
• ✔ Queried claims data set of 300,000+ rows using MySQL Workbench and fed query results directly into R using the package RMySQL for further analysis.

Sep 2020 – Mar 2021 · 6 mos · Chicago, Illinois, United States

• I was brought into the company to leverage Multiple Linear Regression, Boosted Decision Tree, K-Nearest Neighbors, K-Means, Feed Forward Neural Networks, and Multi-Layer Neural Networks in R to provide predictive modeling and machine learning. In this role, I trained and mentored higher management on using R programming.
• SELECTED ACHIEVEMENTS:
• ✔ Wrote complex code in R to automate data preprocessing, data cleaning, predictive modeling and data visualization. R packages used – dplyr, data.table.
• ✔ Developed predictive models in R on datasets with more than 17 million rows.
• ✔ Machine Learning algorithms developed in R – generalized additive models, generalized linear models, neural networks and regression spline. R packages used – gam, glm, nnet, neuralnet.
• ✔ Performed statistical goodness of fit t-tests on claims data after aggregating the data to zipcode resolution using summarize() and group_by(), to compare the difference in means of the modeled data with the observed claims data at a 5% significance level.
• ✔ Developed more than 1000 graphs and bar charts for different construction types and compared different sets of data after data cleaning and data wrangling. The visualizations I developed in R were important in assessing the accuracy and quality of Impact Forecasting’s catastrophe model for the Florida Loss Commission. R packages used – ggplot2

Sep 2016 – Sep 2020 · 4 yrs · Chicago, Illinois, United States

• SELECTED ACHIEVEMENTS:
• ✔ Spatially simulated the largest dataset in the history of Impact Forecasting’s SCS model - 609 million severe convective storm events, and improved model accuracy by developing a revolutionary 5 km by 5 km grid in R. Executed loss calculation for 10 US states in the tornado model portion and wrote code in R for data wrangling, spatial simulation, data visualization, probabilistic modeling, looping and error handling.
• ✔ Cleaned, sorted hurricane data, preprocessed data, and performed linear interpolation. Wrote code in R using the k-nearest neighbors’ algorithm to generate simulations of hurricane track and central pressure, which compared well with historical measurements. Wrote code in R for linear interpolation, data cleaning, predictive modeling, simulation and data visualization.
• ✔ Developed robust predictive models for thunderstorm magnitude by state as a function of path length, path width, latitude, longitude and other parameters using Boosted Decision Tree, Feed Forward Neural Networks, and Multi-Layer Neural Networks in R.
• ✔ Developed robust predictive models for radius to maximum wind speed (Rmax), maximum wind speed (Vmax), and central pressure (Pc) for hurricane events in the North Atlantic Ocean Basin using Multiple Linear Regression. Developed simulations of Rmax against Delta P which compared well with historical measurements of Rmax.
• ✔ Collected and analyzed publicly available data from the US Census Bureau and other sources to develop the US insurance industry exposure database; queried data with >1M rows using Microsoft SQL Server.