Devin Kestell Email & Phone Number

Madison, Wisconsin, United States

Dead machines don't need parts! I lead investigative work modeling machine attrition over time. I then put this work into production providing predictions to millions of machine. In addition, I also predicted customer churn modeling how and when our customers take their business from one dealership to another. With these two pieces of information, our team is better able to predict parts demand at each dealership location. This shortens lead times, increase sales, and lowers required… Show more Dead machines don't need parts! I lead investigative work modeling machine attrition over time. I then put this work into production providing predictions to millions of machine. In addition, I also predicted customer churn modeling how and when our customers take their business from one dealership to another. With these two pieces of information, our team is better able to predict parts demand at each dealership location. This shortens lead times, increase sales, and lowers required stocking levels. For this and other work our team did, we won a 2023 John Deere Presidents Award. Show less

Des Moines, Iowa, United States

A black box model often does not cut it. This is why explainable AI is so important. To help create more explainable AI, I am currently feature engineering on a large data set with around 200 thousand unique features. Condensing these features into a small, manageable and explainable subset is the goal of this project. In this project, I am using end-to-end learning, TF-IDF, clustering and SHAP to provide explanatory lift.

Moline, Illinois, United States

I used a Meta AI Natural Language Processing model to classify four billion dollars of spending into 13 classes. This model ultimately achieved an accuracy of over 95 percent. For this work, our team won a 2021 Vice President's Award.

Madison, Wisconsin, United States

In this role, I taught three different courses for undergraduates: introductory statistics for business, experimental design, and linear regression. It is one thing to teach student's an unapproachable formula like β = (XTX)^−1 X^TY. But it is another thing to take students by the hand and guide them to a deeper, more intuitive understanding of statistics. My student's evaluations of my teaching placed me in the top 5% of statistics TAs in the spring of 2022 semester.

Madison, Wisconsin, United States

Regardless of what they do, any good innovator needs a space for creativity: some call it a sandbox, others might call it a workbench. My role was to help provide the data and visualizations for our clients innovate. To create this data sandbox, I utilized R to scrape data from the web, before cleaning and joining the data. From there, I visualized the data with interactive Shiny apps to give context to the data and allow our clients easier exploration.

Madison, Wisconsin, United States

Finding the right doctor can have a huge impact on both your health and your wallet. Yet, we lack rigorous and encompassing performance ratings for physicians. My goal here was to help bridge this gap.I used SQL to analyze proprietary insurance records and R to collect and analyze external sources of information. At the end of the internship, I presented my methodology and findings to company executives and deployed the results to an internal table.

Oshkosh, Wisconsin, United States

When converting a factory from a legacy system to SAP, a lot of data needs to be created. My role was to help ensure the data was accurate and consistent with internal standards and rules. To help automate this process, I used Excel and SAP.

Master'S Degree, Statistics, 3.8

Bachelor Of Science - Bs, Statistics, Gpa:3.7, In Major Gpa: 3.98

Activities and Societies: Academic Honors: Dean’s List (Fall 2015, Spring 2019)