Sharan Rajamanoharan

Data Scientist @Xcel Energy

Migrated predictive customer behavior models from third-party systems to a custom in-house Databricks pipeline (feature selection, logistic regression, random forests). • Reduced vendor costs by $16M+ while improving forecasting accuracy and ensuring internal model ownership. • Deployed LLM-powered NLP solutions (LLAMA, Claude) to analyze corporate calls and emails; performed EDA on unstructured text for trend detection and automated decision support-improving executive response times. • Developed and launched an A/B testing framework to measure patrol route strategies; integrated clustering methods and Google OR-Tools for optimized routing-boosting efficiency across powerline nodes. • Built interactive data visualization dashboards in Streamlit (hosted on Posit Connect) to analyze five years of community energy use-guiding local energy planning and policy decisions.