Designing algorithms that decide how millions of packages move through large-scale delivery networks — balancing density, cost, speed, and carrier capacity in real time.

• • Combinatorial optimization for package-to-route assignment at national scale
• • Geospatial cost estimation with orders-of-magnitude granularity improvements
• • Real-time decision systems processing millions of packages daily

Building forecasting platforms that manage millions of concurrent prediction models across global regions, combining statistical, deep learning, and ensemble methods to drive operational planning.

• • Multi-horizon forecasting architecture spanning 6 global regions
• • Hierarchical reconciliation across station, region, and network levels
• • Short-horizon models for real-time capacity alignment during demand surges

Applying optimization and machine learning to warehouse floor operations — from task sequencing and makespan minimization to physical storage design.

• • ML-driven task duration prediction combined with integer programming for scheduling
• • Combinatorial optimization for configurable storage unit design
• • Won Top 2 OR Paper Award at Amazon Machine Learning Conference 2022

Game-theoretic modeling of subscription plan design under retail competition — determining when universal free delivery outperforms tiered or paid alternatives.

• • Universal free-delivery subscriptions generate 33.7% more profit vs. paid delivery
• • Published in Production and Operations Management (2024)
• • Analytical framework for competing retailers with heterogeneous consumers

Optimizing train routing, scheduling, and infrastructure health across BNSF Railway's 32,500-mile national network — solving capacity-aware problems at previously intractable scale.

• • Capacity-aware routing and scheduling across 28 states and 3 Canadian provinces
• • Large-scale models (10M+ variables, 20M+ constraints) with sub-second runtime
• • Sensor-based defect detection processing terabytes of real-time data