projects
business initiatives in marketing, data analytics, and optimization — from predictive modeling to portfolio optimization and strategic campaigns
Built a predict-then-optimize pipeline to help policymakers allocate land across conservation, reforestation, and agriculture in Indonesia, Vietnam, and the Philippines. Trained a regression model on FAOSTAT historical data to predict crop yields under different land configurations, then solved a linear program in Julia/JuMP maximizing carbon sequestration subject to a food security floor. Surfaced the Pareto frontier between climate and hunger outcomes to make the policy tradeoff explicit.
Applied advanced marketing strategies to reinvent the approach to marketing at Scentbird, a luxury fragrance subscription company. Performed market research to gather data for customer segmentation, targeting, and positioning. Utilized conjoint analysis to quantify customer preferences, proposed a dynamic pricing model, and designed a two-pronged campaign aimed at improving customer acquisition efficiency and retention.
Combined predictive modeling (LightGBM, SHAP) and mathematical optimization to support real estate investment decisions. Predicted housing prices on the Ames dataset and optimized portfolio allocation under budget and diversification constraints.
Operations-focused engagement with health-tech startup Medikana: worked directly with the company on the "All-Hands Product Push," analyzed their structure and sales pipeline, and recommended changes including a centralized CRM for coordination and scalability.
Researched and designed a community proposal for the City of Cambridge's Participatory Budgeting process to install resident-built nesting boxes across public green spaces. The initiative supports passive pest control, biodiversity, and civic engagement, bringing back birds and community camaraderie to the urban environment.
Developed a mixed-integer program using JuMP and Gurobi to model large-scale hurricane evacuation in Florida. Integrated real-world road, shelter, and population data to minimize unallocated evacuees under capacity and distance constraints. Achieved a 79% improvement over greedy heuristics and identified high-impact infrastructure sites.