Plant Assignment Optimization Using Mixed-Integer Programming | Descriptive & Diagnostic Analytics

Built a Generalized Assignment Problem (mixed-integer linear program) in Python (PuLP/CBC) to optimize plant-order routing across 9,215 real orders and 19 manufacturing plants; found a $146K (0.9%) cost-reduction opportunity while enforcing real capacity, product-eligibility, and contractual (VMI) constraints. Conducted Pareto/ABC concentration, delivery-performance, carrier scorecard, and capacity-utilization analysis using pandas and matplotlib, surfacing operational insights (e.g., 63% of volume concentrated in 5 of 46 customers) beyond the core optimization.

Business Analytics Practicum | Strategy & Consulting – MSBDC | UMass Amherst

Performed EDA on MSBDC client, sales, and trade data across 14+ counties (FY22–FY25); mapped international trade activity by country to identify Europe as the dominant export market for MSBDC-assisted businesses, ahead of Canada and Asia-Pacific. Segmented clients by industry and consultation pattern, uncovering a mismatch: MSBDC directed the most consultation hours toward service-based clients, while product-based businesses showed faster export growth in Europe despite receiving comparatively less support. Recommended reallocating consultation resources toward these higher-performing client segments, giving MSBDC leadership data to strengthen funding and policy conversations with state and federal partners on small business investment.

Supply Chain Analytics | Geospatial Optimization – Amazon Prime Air

Designed a 3-hub drone fulfillment network across 14 Massachusetts counties serving 5M+ residents by applying Center of Gravity and Huff Gravity models to demographic, e-commerce, delivery-distance, package-weight, weather, and bandwidth data to site candidate drone hub locations, validating coordinates against Google Earth. Identified Newton hub as capturing 72.35% of statewide demand (4,926 daily deliveries within a 5-mile radius) versus Worcester (1,068) and Springfield (1,020), informing lead-time and inventory-allocation strategy. Recommended Newton as the primary hub, with Worcester as a secondary option, delivering a phased deployment roadmap and executive memo covering procurement trade-offs, rollout-sequencing cost savings, regulatory constraints (FAA BVLOS), and open risks of drone range/endurance and real-time data availability for leadership to resolve before a full state rollout.

Pricing Intelligence Model – Airbnb (New Orleans)

Preprocessed 7,444+ Airbnb listings using IQR-based outlier detection; engineered 10+ features including amenity count, host tenure, availability, and guest capacity. Identified bathrooms as the strongest price driver (44% feature importance) via an optimized Random Forest model achieving R² = 0.58 and 86.6% of predictions within $100 of actual price. Uncovered neighborhood-level pricing disparities (French Quarter average: $5,660 vs. citywide median: $111), informing host pricing strategy and platform dynamic-pricing recommendations.

Machine Learning | Model Benchmarking – Credit Card Default Prediction

Benchmarked Logistic Regression (AUC 0.772), Classification Tree (AUC 0.760), and Neural Network (AUC 0.696) across 25+ architectures on 10,000 customer records for credit-default prediction. Performed EDA, data cleaning, dummy-variable creation, tree pruning, and lift-chart analysis to optimize model cutoff thresholds (0.26 selected for best F1 = 0.542). Recommended Logistic Regression for production deployment based on superior F1 score and regulatory interpretability, with a three-tier intervention strategy (high / medium / low risk).

Statistical Analysis – Hybrid Work Adoption Among U.S. Employers

Tested the hypothesis that >50% of U.S. companies have permanently adopted hybrid work by 2026, using a systematically constructed sample of 29 companies across industries (IT, Finance, Retail, Manufacturing, Health Services, Hospitality, Government, Education). Observed that the hybrid rate of 62.1% did not reach statistical significance at α = 0.05 (z = 1.30, p = 0.097); 95% Wilson CI [0.44, 0.77] spanned the null boundary. Key takeaway being conclusion is industry conditional, IT industry reached almost 80% hybrid while the other industries were significantly lower.

✅ Verified by Extern · ⏱️ In progress

Amazon Operational Strategy & People Analytics Externship

Imagine shaping how AMAZON — yes, THE Amazon — welcomes and keeps its workforce. Every year, Amazon loses billions because new hires leave before Day 90. You’re going to fix that. In this externship, you’ll web scrape real-world employee reviews from across the web, extract actionable insights, and design persona-driven strategies to help Learning Ambassadors connect smarter, faster, and more humanely. It’s part research, part strategy—and all about learning how people data drives big decisions.