Projects

Multi-Scenario Bus Dispatching and Real-Time Routing Engine

Architected a bus-dispatching engine supporting fixed-route, dynamic ride-pooling, and fusion modes to improve fleet utilization and passenger journey times. Formulated and solved real-time routing models, including open pickup-and-delivery TSP for dynamic cruising and fixed-endpoint TSPP for segment-based route deviations.

A Hybrid Approach to Rideshare Fleet Sizing: Integrating M/M/s Queueing Models with Agent-Based Simulation (Columbia University Evening Shuttle)

A hybrid fleet-sizing study for Columbia's evening shuttle service, combining an analytical M/M/s queueing benchmark with agent-based simulation to evaluate driver staffing, wait time, time in system, and driver utilization under alternative operating configurations.

Chicago Divvy Bike Supply & Station Management

Developed integer and mixed-integer optimization models for bike supply, station selection, and pricing-aware station management in Chicago's Divvy bike-share network.

Eliminating Child Care Deserts in New York State: A Budget-Minimizing Siting and Expansion Model

Formulated an integer optimization framework for deciding where to build new child-care facilities and how to expand existing capacity under ZIP-level coverage, age-specific capacity, and distance separation constraints.

Airline Lunchbox Inventory Optimization

Developed a daily ordering strategy for airline meal preparation, balancing perishable fresh components, reusable snack kits, shortage vouchers, fixed ordering costs, holding costs, and storage limits under uncertain passenger demand.

NYC Last-Mile Trailer Hub Location and Allocation

Formulated a weekly last-mile delivery optimization model for positioning semi-trailer hubs across New York City, allocating neighborhood-level demand, and managing trailer capacity under uncertain delivery cost and unmet-demand penalties.

Oil Tanker Routing and Loading Optimization

Designed a quarterly routing and loading strategy for an oil tanker fleet serving multiple ports, combining vehicle routing, capacity allocation, stochastic travel times, uncertain port demand, and variable oil prices.

Stochastic Warehouse Staff Scheduling

Designed a one-day-ahead staff scheduling model for a 24/7 GPU-refactoring warehouse, accounting for random gig-worker attendance, uncertain productivity, payroll cost, and expensive overflow coverage.

LLM Inference Server Scheduling

Developed a server activation policy for an LLM inference data center, balancing electricity cost, memory overhead, and SLA/congestion penalties under stochastic arrivals and queueing dynamics.

Multi-Echelon Supply Chain Production Optimization

Built a production-planning model for a multi-facility soda supply chain network, setting monthly production quantities under Newsvendor-like profit functions, capacity constraints, inter-facility demand propagation, and stochastic external demand.

Diffusion Models with a Time-Conditioned U-Net on MNIST

Implemented and studied denoising diffusion models with a time-conditioned U-Net architecture for image generation.

Probabilistic Machine Learning and Variational Inference Implementations

Python implementation portfolio covering Bayesian linear regression, hidden Markov models, Gibbs sampling, EM, coordinate-ascent variational inference, variational autoencoders, Poisson matrix factorization with mean-field variational inference, variable elimination, and forward-backward inference.

Computer Vision for Quality Inspection of Manufacturing Product

Applied computer vision methods to manufacturing quality inspection, combining texture features, convolutional neural networks, and model interpretation.

Online Twitter Bot Detection

NLP project for identifying automated social-media accounts using neural networks and transformer-based modeling.

Exploring Card Fraud Transaction

Applied supervised learning methods, including decision trees, k-nearest neighbors, logistic regression, and related classifiers, to card-fraud transaction detection.

Introduction to Big Data Systems

Built data-system projects using PyArrow, Docker, HDFS, Spark, HBase, Cassandra, and Kafka Streams.

Database Management Systems (DBMS): Design and Implementation

Completed SQL query projects and C++ implementation work for core database internals, including buffer management and join algorithms.

AdTracking Fraud Detection

High-performance analytics project demonstrating large-scale data processing and model experimentation using UW-Madison's HTCondor-based high-performance computing environment.

A Spatial Temporal Visualization Tool of Climate-Demographic Interactions and Built Environments

Built a geospatial visualization tool connecting climate projections, population, land use, housing density, and city boundaries across long-horizon RCP 4.5 / RCP 8.5 scenarios.

Interactive prototype → A sample panel rendered on New York State data.

Statistical Experiment Design & Analysis

Experimental design and statistical analysis project using ANOVA and Fisher's LSD procedure.

Unpacking the Paradox of Profit Margins: An Analysis of Market Concentration and Economic Dynamics in the U.S. Meat Processing Industry

Economic analysis of market concentration and profit margins in the U.S. meat processing industry.

Presentation →

Analysis of NHL Player Statistics

Introductory statistical analysis and visualization project using NHL player-performance data.

Examining the Relationship Between Socioeconomic Factors and HIV Prevalence in the United States

Applied econometrics and data analysis project examining associations between socioeconomic variables and HIV prevalence.