Built an end-to-end retail sales forecasting pipeline on 1.01M rows of historical data from 1,115 Rossmann stores, comparing LSTM and XGBoost on identical feature sets. Engineered lag features, rolling statistics, promotions and holiday variables per-store to eliminate cross-store data leakage. Tuned XGBoost achieved MAE 570, RMSE 799, R² 0.928, while the LSTM hit 12.27% MAPE — a 66.4% improvement over the naive baseline. Deployed as an interactive Streamlit dashboard with store-level forecast simulation, a promo toggle, and model comparison across all 1,115 stores.

Built an XGBoost-based churn prediction and segmentation system on airline loyalty data covering 16,737 customer records. Engineered features from recency, flight frequency, redemption ratio and seasonal engagement trends, achieving 97% classification accuracy and a 0.97 ROC-AUC on imbalanced data through threshold tuning. Added KMeans clustering for customer segmentation and shipped an interactive Streamlit dashboard for churn analysis and retention strategy reporting.

Segmented 2,200+ customers into four behavioral groups using K-means clustering on engineered demographic and spending features, then built an interactive Streamlit dashboard translating cluster insights into targeted marketing recommendations..

Built an interactive Excel dashboard to analyze retail sales performance using Pivot Tables, Pivot Charts, and Slicers. The dashboard enables dynamic exploration of sales trends, customer demographics, regional performance, and sales channels while demonstrating data cleaning, visualization, and dashboard design skills..