SURAJ KUMAR JHA

Defined modeling scope for micro-segmentation and assortment/quantity prediction by translating business objectives into actionable problem statements. Built geospatial/transactional data pipelines, implemented HDBSCAN store clustering, and developed monthly XGBoost forecasting models with accuracy/lift-based acceptance thresholds and executive-ready dashboards.

Implemented a Two-Tower personalization model that connects customer store location and browsing history to deliver omni-channel product recommendations for Lowe’s eCommerce. Built multiple linear regression for bestseller ranking prediction and developed sales forecasting using a composite of forecasters (GBDT, GLM, and Constant) to improve planning and inventory decisions.

Developed a data-driven inventory management approach using demand prediction and probabilistic methods to support ordering decisions while reducing stockouts/overstock. Created sales forecasting using a composite set of forecasters to improve the reliability of predictions for operational planning.

Built a recommendation engine for sales representatives by identifying predictive levers to improve beer market share and translating scenarios into stakeholder-ready opportunities. Integrated large-scale internal and external variables, trained Random Forest models over massive combinations, and developed retailer segmentation using K-means clustering; also implemented a customer identification p

Developed and optimized an insurance claims fraud detection model using Python-based NLP and text analytics. Implemented preprocessing with NLTK and regex/pattern matching, then built and trained classification models with scikit-learn.

Completed a Bachelor of Technology program at Vellore Institute of Technology from July 2010 to May 2014.