Kushagra Verma

I’m a B.Tech student focused on Data Science and Artificial Intelligence, and I love turning messy data into clear decisions. My work centers on end-to-end analytics: shaping the data, building dependable metrics, and delivering interactive insights people can trust.

In my churn analysis project, I engineered a T-SQL feature pipeline on 10,000+ customer records using temp tables and CASE WHEN segmentation to create consistent categorical buckets. I built a multi-layer Power BI dashboard with dynamic DAX-driven titles and field-parameter toggles, and I surfaced $186M in lost deposits tied to churn patterns—highlighting Germany as the highest-risk region.

I also enjoy discovering storytelling in data, from cleaning and parsing unstructured datasets to designing analysis-ready visuals. In my Netflix catalog study, I processed 8,800+ titles, handled date anomalies, and created a custom dark-themed visualization engine across 9 charts to reveal content pivots and micro-trends.

On the modeling side, I built GlucoseIQ as a full-stack Diabetes Risk Predictor by improving ROC-AUC from 0.72 to 0.81 with GridSearchCV and deploying it via a Flask API with a React/Vite frontend. I’m also contributing to research through an IEEE paper on 5G handover prediction under review, where I work on preprocessing, feature engineering, and model benchmarking.