At Colaberry Inc, I led a team of four data scientists in modernizing enterprise demand forecasting. We replaced labor-intensive processes with automated machine-learning workflows that generated multi-million-dollar business savings.
I built Quantile Regression and OLS models to forecast hourly demand across a six-state operational network, achieving 0.23 Pinball Loss, approximately 25% SMAPE, and approximately 15% MAPE.
I also developed a customer propensity model using Random Forest classification and a computer-vision solution using OpenCV-based processing and Naïve Bayes classification. The computer-vision solution achieved 98.1% validation accuracy and reduced reliance on manual observation.
I designed a workforce optimization engine using K-Means and Google OR-Tools that reduced average travel time from 15 to 12 hours. In a personal project, I built a Python RAG Q&A tool for S&P 500 SEC filings, with citation-validated summaries for investment research.
