Sunjae Bae

I am an Assistant Professor of Surgery and Population Health at NYU Grossman School of Medicine.

I study clinical outcomes and treatment decisions in kidney disease and transplantation using nationwide data from Epic Cosmos, SRTR, and USRDS. My research emphasizes heterogeneous treatment effects and causal machine learning, with agentic AI for clinical research as an emerging focus.

I lead NIH/NIDDK-funded projects (R01DK144783, K01DK132490) and serve as a co-investigator on other NIH-funded studies. Previously, I was an instructor at Johns Hopkins, following four years in clinical practice.

Research

Transplant Epidemiology

I study treatment effects, individualized outcome prediction, and organ allocation in kidney transplantation. I use data from SRTR, USRDS, and Epic Cosmos to examine who is most likely to benefit from or be harmed by immunosuppression and other treatments, and how the associations between established risk factors and transplant outcomes vary with donor and recipient characteristics.

Causal Machine Learning

I develop and apply causal inference and machine learning methods to estimate treatment effects from real-world clinical data. One priority is a unified workflow for estimating ATE/ATO/HTE. I’m also exploring flexible ways to represent patient profiles and clinical events from large, longitudinal EHR datasets for causal inference.

Agentic Clinical Research

I’m exploring how AI agents can support and accelerate clinical research while ensuring rigorous study design, making analytical decisions transparent, and protecting research data/PHI. My short-term goal is to train a lightweight model that can build and run a complete, reproducible data analysis pipeline entirely on a MacBook Pro or comparable consumer hardware.

Selected papers

Full bibliography

Early steroid withdrawal in first and repeat transplants

Bae et al., Nephrology Dialysis Transplantation, 2025.

Early steroid withdrawal was associated with slightly lower mortality but slightly higher graft-failure risk in first-transplant recipients. In repeat-transplant recipients, rejection and graft-failure risks were substantially higher, without an apparent mortality benefit.

Effect of early steroid withdrawal, first transplant versus retransplant Ratios for acute rejection, graft failure and death. First transplants: 1.04, 1.09 and 0.93. Retransplants: 1.42, 1.24 and 1.01. First transplant Retransplant Acute rejection Graft failure Death 0.91.01.21.41.6 ← better with withdrawal worse →

Work with me

Contact me via my NYU email.

Trainees

Want to apply causal ML and agentic AI to large datasets (Epic Cosmos, SRTR, Medicare claims) to answer open clinical questions?

Open to all stages, from undergrad and master's/PhD to med school, residency, and fellowship. Join for a summer sprint, a side quest during school, or a full-time postdoc or analyst role.

Email me with:

  1. Your CV.
  2. Your current training or role and intended timeline.
  3. Why you want to work with me and/or a clinical or analytical question that interests you.
  4. Prior work, if any (papers, code, GitHub repositories, etc.).

Collaborators

I welcome collaborations with academic, clinical, and industry teams. If you’re interested in comparative effectiveness or other real-world evidence studies, or in evaluating a clinical prediction model or an AI tool for research, please reach out.