Stefan Faulkner
Optimization & Machine Learning for Decision-Making in Healthcare and Logistics
I am a Ph.D. student in Machine Learning at the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech, advised by Prof. David Goldsman and co-advised by Prof. Edwin Romeijn. I am a 2026–27 George Fellow in health systems.
My research uses mathematical optimization, simulation, and deep learning to study cancer treatment scheduling, screening resource allocation in low-resource settings, and delivery delays in supply chains. I work with clinical, global-health, and industry partners on these problems, with a focus on helping them make decisions under uncertainty.
Originally from Kingston, Jamaica, I studied Computer Science and Mathematics at the University of the West Indies, Mona, and worked as a data scientist at National Commercial Bank before coming to Georgia Tech. I have also spent two summers as a Research Scientist intern at Chewy, working on optimization and machine learning problems in inventory and logistics.
Research
Submitted Paper
Uncertainty-Aware Delivery Delay Duration Prediction via Multi-Task Deep Learning
Under review at Expert Systems with Applications
Develops a multi-task deep learning model that predicts whether a shipment will be late and by how much, designed for highly imbalanced data where delays are rare. Trained end-to-end on 10M+ industrial shipment records, it provides calibrated prediction intervals and reduces delayed-shipment error by 41–64% over single-step tree-based baselines and 15–35% over two-step pipelines.
Working Papers
Personalized Treatment Scheduling for Prostate Cancer via ODE-Constrained Mixed-Integer Programming
Develops a multi-objective, ODE-constrained mixed-integer nonlinear programming framework for personalizing intermittent hormone therapy in prostate cancer. Built on a validated tumor-dynamics model, it explores the trade-offs between treatment burden and PSA control, with the aim of quantifying how much treatment personalized schedules can spare relative to standard fixed protocols.
Optimal Cervical Cancer Screening Allocation for Women Living with HIV in Zambia: A Markov–LP Framework
Develops a simulation–optimization framework for allocating cervical cancer screening among women living with HIV in a resource-constrained setting. A Markov cohort model of disease progression feeds a linear program that selects screening protocols and intervals under facility budget and capacity limits, revealing which resources constrain early detection and the value of investing in each, with probabilistic sensitivity analysis used to assess the robustness of these findings. This work is done in collaboration with CIDRZ, San Francisco State University, and Kanyama General Hospital.
Modeling Customer Arrivals from Airline Booking Data: An Application to Revenue Management Simulation
Examines how revenue management simulators generate customer booking requests over the booking horizon. Using large-scale airline booking data, it evaluates common demand-generation methods against observed booking behavior and develops an arrival process that better reflects it, with the aim of assessing its impact on simulated revenue management performance.
News
- Nov 2026 UpcomingPresenting customer arrival-process research at the ATL@GT meeting, Herndon, VA (Nov 17–18).
- Nov 2026 UpcomingTalk on personalized prostate cancer treatment scheduling at the INFORMS Annual Meeting, San Francisco, CA (Nov 1–4).
- Fall 2026 Named a George Fellow (2026–27) for health-systems research at Georgia Tech ISyE.
- Summer 2026 Returned to Chewy as a Research Scientist II intern, joining the Inventory Technology & Science team to build graph-based reinforcement learning policies for fulfillment-center inventory transfers.
- May 2026 Presented personalized prostate cancer treatment optimization at the IISE Annual Conference & Expo, Arlington, TX.
- Spring 2026 Awarded the Coca-Cola Foundation Fellowship at Georgia Tech.
- Feb 2026 Launched a cervical cancer screening evaluation with CIDRZ, San Francisco State University, and Kanyama General Hospital in Lusaka, Zambia.
- Feb 2026 Delivery delay prediction paper posted on arXiv.
- Jan 2026 Began research on airline customer arrival processes with a team of graduate students at Georgia Tech's Air Transportation Lab.
- Oct 2025 Presented delivery delay prediction at the INFORMS Annual Meeting, Atlanta, GA, and served as a student volunteer.
- Summer 2025 Research Scientist II intern on Chewy's Supply Chain Replenishment Science team, building an LLM + optimization multi-agent system for purchase-order planning.
- Jun 2025 Presented at the IISE Annual Conference, Atlanta, GA (AI4OPT highlight).
- Summer 2024 Seth Bonder Teaching Fellow at AI4OPT, teaching Python, data science, and deep learning to high school students.
- Apr 2024 Featured in Georgia Tech ISyE news: “Transforming Package Delivery with AI Solutions.”
- Apr 2023 Featured by Georgia Tech Professional Education: from Jamaica to a Ph.D. in Machine Learning.
- Aug 2022 Started my Ph.D. in Machine Learning at Georgia Tech.
Presentations
- Modeling Customer Arrivals from Airline Booking Data ATL@GT Meeting, Herndon, VANov 2026
- Personalized Treatment Scheduling for Prostate Cancer via ODE-Constrained Mixed-Integer Programming INFORMS Annual Meeting, San Francisco, CANov 2026
- Personalized Treatment Scheduling for Prostate Cancer via ODE-Constrained Mixed-Integer Programming IISE Annual Conference & Expo, Arlington, TXMay 2026
- Data-Driven Delivery Delay Prediction Using a Deep Learning Framework INFORMS Annual Meeting, Atlanta, GAOct 2025
- Data-Driven Delivery Delay Prediction Using a Deep Learning Framework IISE Annual Conference & Expo, Atlanta, GAJun 2025
Experience
- Research Scientist II Intern, Chewy Inventory Technology & Science: graph neural network RL for inventory transfersSummer 2026
- Research Scientist II Intern, Chewy Supply Chain Replenishment Science: LLM + mixed-integer programming agentsSummer 2025
- Graduate Research Assistant, Georgia Tech ISyE Optimization for healthcare decision-making; previously machine learning for delivery delay prediction with AI4OPT and Kinaxis (2023–2025)2022 – present
- Data Scientist, National Commercial Bank Jamaica Enterprise Information Management Unit2019 – 2022
Education
- Ph.D., Machine LearningGeorgia Tech (ISyE)2022 – Spring 2027 (expected)
- M.S., Computational Science & EngineeringGeorgia Tech2022 – 2024
- M.S., Computational AnalyticsGeorgia Tech2020 – 2022
- B.S., Computer Science & MathematicsUniversity of the West Indies, Mona2016 – 2019
Honors & Awards
- George FellowshipGeorgia Tech ISyE2026–27
- Coca-Cola Foundation FellowshipGeorgia Tech2026
- Seth Bonder Teaching FellowshipAI4OPT2024
- ISyE FellowshipGeorgia Tech2022–23
- ISyE Graduate Travel AwardGeorgia Tech2022
Teaching & Service
- Seth Bonder Teaching Fellow, AI4OPT Taught Python, data science, and deep learning to high school students from underrepresented backgroundsSummer 2024
- Graduate Teaching Assistant, Georgia Tech ISyE Optimization and statistics courses2022 – 2023
- Student Volunteer, INFORMS Annual Meeting Atlanta, GA2025