Recent Talks: I presented the DELPHI model and the Janssen collaboration at the 2021 INFORMS Annual Meeting in Session VWD12/MB22 (Virtual/In-person) and also VSA84/TE39 for the Doing Good with Good OR competition. On behalf of Prof. Bertsimas, I discussed our work on optimizing COVID-19 vaccine allocation in Session VTC72. My co-author Vassilis Digalakis presented our research on a scalable algorithm for slowly-varying sparse regression in Session VSA07/TC14.
A selection of my research and industry projects are explained in the various sections on the left. I have broadly separated my projects into areas/industries that I have felt they are chiefly concerned with. For methodologies that I feel are broadly applicable to different industries, I have put them under the General category.
As explained on my philosophy page, I care about utilizing the interdisciplinary tools that I know (statistics, optimization, and analytics) to create impact. I work on both methodology and applications to achieve this goal, as briefly outlined below:
My methodological research focus on developing new tools for prediction and evaluation in large-scale data analytics/machine learning. They mostly consists of two streams:
1. Scalable Algorithms for Data-Driven Prediction - This stream of work focuses on developing exact and approximate algorithms for NP-hard problems in machine learning and statistics that are applicable to modern data sizes, including matrix completion and sparse regression.
2. Evaluation/Inference for Machine Learning - Another key focus of my research centers around evaluating machine learning methods. This includes both developing new algorithms for consistent inference under observational data, and also creating evaluation metrics for rigorous evaluation of black-box prescription methods.
On applications, I have been focused on healthcare and epidemiology. During the COVID-19 pandemic, I have worked on topics ranging from advising governments on non-pharmaceutical restrictions to accelerating the Phase III trial for the J&J vaccine, contributing to saving thousands of lives around the world. Separately, I also have a continuous stream of work collaborating with Boston Children's hospital working on applying my methodological research to create and evaluate personalized treatment rules for pediatric patients.
The full list of my papers and preprints are as followed, organized by topic:
Scalable Algorithms for Data-Driven Prediction
Fast Exact Matrix Completion: A Unified Optimization Framework for Matrix Completion (with Bertsimas, D.). Journal of Machine Learning Research, 21(231), 1-43 (2020).
Stochastic Cutting Planes for Data-Driven Optimization (with Bertsimas, D.). arXiv preprint arXiv:2103.02506 (2021). Submitted to Mathematical Programming.
Slowly Varying Regression Under Sparsity (with Bertsimas, D., Diaglakis, Lami, O. S). arXiv preprint arXiv:2102.10773 (2021). Submitted to Journal of Machine Learning Research.
Scalable Holistic Linear Regression (with Bertsimas, D.). Operations Research Letters (2020).
Interpretable Matrix Completion: A Discrete Optimization Approach (with Bertsimas, D.). arXiv preprint arXiv:1812.06647 (2020). Submitted to Machine Learning.
Finalist, 2020 Mixed Integer Programming (MIP) Workshop Best Student Poster Competition
Evaluation/Inference for Machine Learning
Experimental Evaluation of Individualized Treatment Rules (with Imai, K.). Journal of the American Statistical Association (2021): 1-41.
Pricing for Heterogeneous Products: Analytics for Ticket Reselling (with Alley, M., Biggs, M., Hariss, R., Herrmann, C., & Perakis, G.). Available at SSRN 3360622. (2019). Accepted at Manufacturing & Service Operations Management.
- Finalist, 2018 M&SOM Practice-Based Research Competition
Data-Driven Vaccine Development with Janssen (with Bertsimas, D., Xu, J., Khan, N.) Working Paper. To Submit to Nature.
- Finalist, 2021 INFORMS Doing Good with Good OR Competition
- Semi-Finalist, 2021-2022 INFORMS Edelman Competition (Finalist Selection Pending)
Forecasting COVID-19 and Analyzing the Effect of Government Interventions (with Bouardi, H. T., Lami, O. S., Trikalinos, T. A., Trichakis, N. K., & Bertsimas, D.) medRxiv (2020). Minor Revision at Operations Research.
- Winner, 2021 INFORMS Analytics Society Innovative Applications in Analytics Award
From predictions to prescriptions: A data-driven response to COVID-19 (with Bertsimas, D., Boussioux, L., Wright, R. C., Delarue, A., Digalakis Jr, V., Jacquillat, A., ... & Nohadani, O.). arXiv preprint arXiv:2006.16509. Accepted at Health Care Management Science.
- Winner, 2020 INFORMS Health Applications Society Pierskalla Best Paper Award
Selecting Children with VUR Who are Most Likely to Benefit from Antibiotic Prophylaxis: Application of Machine Learning to RIVUR (with Wang, H. H. S., Li, M., Bertsimas, D., Estrada, C., & Caleb, N.). The Journal of Urology, 10-1097 (2020).
Targeted Workup after Initial Febrile Urinary Tract Infection: Using a Novel Machine Learning Model to Identify Children Most Likely to Benefit from Voiding Cystourethrogram (part of Advanced Analytics Group of Pediatric Urology and ORC Personalized Medicine Group). The Journal of Urology, 202(1), 144-152 (2019).
Prescriptive Analytics for Reducing 30-day Hospital Readmissions after General Surgery (with Bertsimas, D., Paschalidis, I. C., & Wang, T.). PloS one, 15(9), e0238118 (2020).
Where to locate COVID-19 mass vaccination facilities? (with Bertsimas, D., Digalakis, V., Jacquillat, A., & Previero, A.). Naval Research Logistics (NRL) (2021).
Ensemble Forecasts of Coronavirus Disease 2019 (COVID-19) in the US (with Ray, E. L., Wattanachit, N., Niemi, J., Kanji, A. H., House, K., Cramer, E. Y., ... & COVID-19 Forecast Hub Consortium.). MedRXiv (2020).
Short-term forecasting of COVID-19 in Germany and Poland during the second wave – a preregistered study (with Bracher, J., Wolffram, D., Deuschel, J., Gorgen, J., Ketterer, J.L., et al). MedRXiv (2021). Accepted at Nature Communications.
Application of Duration-of-Stay Storage Assignment with Deep Neural Networks under Uncertainty (with Wolf, E., & Wintz, D.). CoRR (2020).
- Spotlight, 2020 International Conference on Learning Representations (ICLR)
A Hierarchy of Graph Neural Networks Based on Learnable Local Features (with Dong, M., Zhou, J., & Rush, A. M. ). arXiv preprint arXiv:1911.05256 (2019).