Hi, I'm Roni
A mathematician at heart and a data scientist by trade. I turn messy data into models people can use, and complicated ideas into something anyone can follow. When I'm not doing that, you'll find me out with my camera!
A little about me
I recently finished my master's in Applied Data Science at East Tennessee State University, where I also worked as a graduate statistical consultant helping researchers design studies and make sense of their data. My projects cover healthcare, retail, education, and bioenergy. Some come from collaborations with Oak Ridge National Laboratory, Eastman Chemical Company, and the Niswonger Foundation; others are personal projects, like fraud detection, sales forecasting, and multivariate health analysis, plus thesis research on synthetic time series. I hold my statistics to a high standard. And I hold my results to a simpler one: they have to make sense to the people who use them.
Putting data to work
Data Science Practicum · Oak Ridge National Laboratory
Oak Ridge studies poplar trees as a source of renewable biofuel, and breeding better ones is slow because you normally have to grow a plant to full size before you know how it turned out. I built an end-to-end Python pipeline over multi-modal images of 446 poplar plants that predicted each plant's final biomass (Lasso cross-validated R² = 0.74) and, using nothing more than ordinary RGB cameras, could flag the top performers as early as day 13 to 14, so researchers can spot promising plants weeks sooner and cheaper than before.
Data Science Practicum · Niswonger Foundation
The Niswonger Foundation helps students across our region get to and through college, and its advisors had three years of notes scattered across inconsistent records that were hard to learn anything from. I reorganized all of it into a clean, normalized 8-table database (95 students and 1,456 advising interactions) and built a Python ETL pipeline that even recovered 180+ case notes that had effectively been hidden in the mess. From there I ran an exploratory analysis with 21 visualizations, turning years of loose paperwork into a picture the foundation could actually use.
Data Science Practicum · Eastman Chemical Company
Eastman's technical staff write up call reports after meeting with customers, but the free-text notes vary a lot in how useful they are. I built a natural-language-processing pipeline that scores the quality of those reports automatically, using LDA topic modeling and Word2Vec to tell substantive, relevant content apart from filler, so the company can find its most valuable notes without reading every one by hand.
Sharing what I know
Graduate Statistical Consultant · Statistical Consulting Lab, ETSU
ETSU's Statistical Consulting Lab offers free statistics help to faculty, staff, and graduate students across the university, and I was one of the consultants people came to. Over the year I supported 30+ research projects from fields as varied as nursing, biology, and education, guiding researchers through the whole process, from designing a study or survey to collecting, organizing, and finally analyzing their data. I ran the analyses in R, SAS, Python, and Excel, and just as importantly, translated the results into plain language for people who don't speak statistics. I also led short workshops and one-on-one training sessions, usually over a two-to-four-week stretch with each client.
Graduate Teaching Assistant · Dept. of Mathematics & Statistics, ETSU
For two years I taught in ETSU's Department of Mathematics and Statistics. One year I was the Instructor of Record for Precalculus (Algebra), running my own section of 80+ students, from lectures and assessments to office-hours support. The other year I served as a teaching assistant for Probability and Statistics, helping with instruction and grading for a course of 200+ students.
Things I've worked on
Credit Card Fraud Detection
Catching fraudulent transactions when only 0.17% are actually fraud. I compared Logistic Regression, Random Forest, and XGBoost, used SMOTE to handle the imbalance, and SHAP to explain why each call was made.
View on GitHub →Cardiometabolic Risk Analysis
Digging into 10,616 NHANES health records to find which biomarkers drive cardiometabolic risk, using PCA, Hotelling's T², and MANOVA.
View on GitHub →Superstore Sales Forecasting
Forecasting future furniture sales from years of retail history, using SARIMA (SAS PROC ARIMA) with Python for the data prep.
View on GitHub →Reading Proficiency vs. Library Funding
Asking whether library funding, staffing, and demographics can predict how well a school's students read, using Lasso, Ridge, Random Forest, and XGBoost (test R² = 0.74).
View on GitHub →ORNL Poplar Phenotyping
Spotting the best biofuel poplars early from camera images alone, with growth-curve modeling, heritability analysis, and predictive ML on ORNL Advanced Plant Phenotyping Lab data.
View on GitHub →DGAN + Soft-DTW Post-Processing
Making AI-generated synthetic time series more realistic, with a post-processing framework using Soft-DTW alignment and ARMA-based statistical evaluation (my master's thesis).
View on GitHub →How I got here
M.S. Applied Data Science
East Tennessee State University
May 2026 · Johnson City, TN
M.S. Mathematical Sciences
East Tennessee State University
Dec 2025 · Johnson City, TN
B.S. Mathematics
University of Dhaka
Dec 2021 · Dhaka, Bangladesh
Through my lens
Let's talk
Have a project, a job opportunity, or just want to say hi?