-Enhanced and maintained a TV program-schedule optimizer which maximizes expected product sales subject to TV scheduling constraints using mixed-integer programming (Gurobi). Implemented dynamic constraint relaxation to handle infeasible scheduling problems. Worked closely with QVC and HSN TV programming teams to update and improve implementation of business logic.-Performed a multivariate analysis of 100+ macroeconomic and internal factors to identify drivers of QVC and HSN sales. Created a feature selection and modeling pipeline to automate the analysis and enable running it across hundreds of scenarios (depending on product division, type of sales, time range, etc.).-Developed an optimizer which allocates TV airtime minutes to product categories a year ahead of time, maximizing expected sales subject to business constraints while minimizing risk, using quadratic programming (Cvxopt). Trained Prophet time series models for each product category to predict both expected value and variance of sales.

-Designed a TV ad placement optimizer using mixed-integer programming which maximizes target audience reach subject to business constraints on the overall ad plan-Automated and parallelized the generation of millions of TV viewership predictions-Designed a method of survey weighting using linear programming and Wasserstein distance-Designed a method of look-alike modeling based on stratified sampling in SQL and solving a linear-sum-assignment problem within strata-Modeled conversion propensity for a given product based on historical usage of related products