Research

Research Interests

My research is at the intersection of precision medicine, causal inference, and sequential decision-making, with applications in healthcare. I am broadly interested in problems where rigorous statistical thinking can directly inform policy and intervention design.

Current projects include work on adaptive treatment policies using online learning in precision medicine, and causal inference with multiple instruments, particularly over-identified instrumental variable settings — I’d be happy to discuss either in more depth. Feel free to reach out, and see the Recent News section on the homepage for the latest updates!

Submitted Papers

2. Proactive Inpatient Bed Requests for Emergency Department Admissions, joint work with Qian Cheng, Nilay Tanik Argon, and Serhan Ziya — submitted to Manufacturing & Service Operations Management.

   arXiv ↗ SSRN ↗

   Problem definition: Emergency department (ED) boarding occurs when patients admitted to the hospital remain in the ED while waiting for inpatient beds. Boarding is viewed as a major driver of ED crowding and has been associated with poor patient outcomes. We propose a framework to help EDs reduce boarding time and total length of stay by using information about current ED patients and hospital bed availability to proactively request inpatient beds before admission decisions are finalized.

   Methodology/results: We formulate the problem as a Markov decision process in which predictions of each patient's probability of hospital admission and time to disposition are aggregated to guide early inpatient bed requests. This formulation leads to three data-driven policies based on approximate dynamic programming, reinforcement learning, and a newsvendor-type approach. Using a simulation model based on data from a large ED, we evaluate these policies across a wide range of settings. The simulation study shows that proactive aggregate bed requests can reduce average boarding times for admitted patients by 30–70% and average length of stay for all ED patients by 6–15%, while creating only modest idle time for prepared inpatient beds. The newsvendor heuristic provides the most attractive tradeoff between ED performance and inpatient bed idle time, whereas the reinforcement learning-based heuristic produces smoother bed-request patterns when stability in downstream hospital processes is especially important.

   Managerial implications: Our work shows how EDs can use prediction tools to make proactive, ED-level bed-request decisions that improve ED operations while helping managers balance reductions in ED delays against limited idle time for prepared inpatient beds. Our findings also illustrate the value of evaluating both simple myopic heuristics and more sophisticated reinforcement learning-based approaches in this operational problem, since each can offer distinct advantages depending on the performance measures and implementation constraints most important to managers.

   @unpublished{cheng2026proactivebeds,
     author        = {Cheng, Qian and Argon, Nilay Tanik and Ganesaraman, Aniruddhan and Ziya, Serhan},
     title         = {Proactive Inpatient Bed Requests for Emergency Department Admissions},
     note          = {Submitted to Manufacturing \& Service Operations Management},
     year          = {2026},
     eprint        = {2607.15432},
     archivePrefix = {arXiv},
     url           = {https://arxiv.org/abs/2607.15432}
   }

3. Data Driven Block Replacement Scheduling, joint work with Vidyadhar Kulkarni — submitted to Operations Research.

   arXiv ↗ SSRN ↗

   We develop data-driven algorithms for maintaining N independent identical machines under a block replacement policy, in which each machine is replaced upon failure and all machines are jointly replaced at regular intervals of length k. The goal is to learn the cost-minimizing interval k* from operational data when the lifetime distribution is unknown. At each decision epoch, the operator selects k ∈ {1, 2, …, K}, observes the resulting failure history (a mixture of complete and right-censored lifetimes) and incurs a per-unit-time cost governed by the renewal function. We formulate this as a stochastic multi-armed bandit and propose Hoeffding- and Bernstein-based lower-confidence-bound algorithms achieving O(K log T) regret, matching the Lai–Robbins lower bound. Exploiting a nested observation property unique to block replacement, correlated variants attain O((K - k*) log T) regret and require only O(1) direct pulls of suboptimal arms k < k*. A complementary Kaplan–Meier renewal algorithm estimates the lifetime distribution nonparametrically from censored data, achieving almost-sure policy consistency and empirically near-zero incremental regret at long horizons. We additionally analyze two average-cost MDPs: a time-elapsed formulation establishing that block replacement is optimal within its policy class for any lifetime distribution, and an age-vector formulation proving a monotone threshold structure under increasing failure rate distributions and providing a gold-standard cost benchmark. Numerical experiments confirm the theoretical ordering and reveal structural cost gaps between optimal block and age-dependent replacement.

   Code for replication is publicly available at this GitHub repository.

   @unpublished{ganesaraman2026blockreplacement,
     author        = {Ganesaraman, Aniruddhan and Kulkarni, Vidyadhar},
     title         = {Data Driven Block Replacement Scheduling},
     note          = {Submitted to Operations Research},
     year          = {2026},
     eprint        = {2607.15229},
     archivePrefix = {arXiv},
     url           = {https://arxiv.org/abs/2607.15229}
   }

Course Projects in UNC

Course projects have been an important part of my research development at UNC. They have allowed me to collaborate, engage seriously with open problems, replicate and extend results from the literature, and explore areas adjacent to my dissertation work.

Semester	Course	Project Title	Materials
Spring 2026	HPM 883 (Causal Machine Learning)	Benchmarking Estimators of Local Average Treatment Effect	GitHub Repo
Fall 2025	BIOS 777 (Precision Medicine)	Deep RL for Personalized Treatment Recommendation (Main Paper)	Report | Slides
Fall 2025	STOR 743 (Reinforcement Learning and MDP)	Deep RL for Personalized Treatment Recommendation (Main Paper)	Report | Slides
Fall 2025	STOR 664 (Applied Statistics I)	NFL Resource Allocations	GitHub Repo
Spring 2026	STOR 672 (Simulations)	Possible causes for increased boarding times	Report | Slides
Fall 2024	STOR 634 (Probability I)	Bayesian NPI of Topic Hierarchies	Report

Conferences & Workshops Attended

Early in my PhD, I have made a deliberate effort to attend conferences and workshops across a variety of areas. I find that broad exposure to different communities and methodologies helps me find my footing and think more creatively about my own work.

Conference / Workshop	Location	Date	Notes
INFORMS Annual Meeting	San Francisco, CA	Nov 2026	Upcoming — Invited speaker, MSOM Service Operations session honoring Prof. Vidyadhar Kulkarni’s retirement; presenting Data Driven Block Replacement Scheduling with Vidyadhar Kulkarni
INFORMS Simulation Society Research Workshop	Raleigh, NC	Jul 2026	Presented poster on Proactive Inpatient Bed Requests for Emergency Department Admissions with Qian Cheng, Nilay Tanik Argon, and Serhan Ziya [Poster]
INFORMS Healthcare Conference	Raleigh, NC	Jul 2026	Presented talk on Proactive Inpatient Bed Requests for Emergency Department Admissions with Qian Cheng, Nilay Tanik Argon, and Serhan Ziya [Slides]
American Causal Inference Conference (ACIC) 2026	Salt Lake City, UT	May 2026	Presented latebreaker poster Estimation of a Common Local Average Treatment Effect with Multiple Instruments with Patrick Lopatto and P. M. Aronow; awarded Honorable Mention in the Tom Ten Have poster competition
Conference on Extreme Value Analysis (EVA 2025)	Chapel Hill, NC	Jun 2025	Volunteered; first exposure to the breadth of extreme value theory and its applications across climate science, public health, and finance
DAIR³	Jackson, MS	May 2025	NIH-supported weeklong bootcamp on rigorous and reproducible biomedical data science; held at Jackson State University
Mini-course by Prof. Remco van der Hofstad	Chapel Hill, NC	Oct 2024	Short course on Dynamic Processes on Networks, hosted by the STOR department at UNC
Advanced Instructional School in Stochastic Processes (AISSP)	Bhubaneswar, India	Jul 2023	Intensive three-week school on Brownian motion, martingales, and measure theory; held at NISER Bhubaneswar [Notes & Problem Sets]