Research

My current research is centered on Multi-Agent Reinforcement Learning, Multi-Agent Systems, and Multimodal AI. I am interested in how learning and decision systems behave when teams change, environments shift, language changes, or AI systems must interact with physical processes.

Multi-Agent Reinforcement Learning Multi-Agent Systems Multimodal AI Robot Learning

Multi-agent learning

Coordination, credit assignment, and robustness

I study cooperative multi-agent learning, with an emphasis on how agents coordinate, how credit should be assigned when team composition changes, and whether apparently permutation-robust policies can still collapse to undesirable behaviors.

Turnover-Orthogonal Credit Assignment for Open-Team Multi-Agent Reinforcement Learning. Preprint, 2026.
Permutation Robustness Is Not Enough: Action Collapse in Multi-Agent Transformer Policies. Preprint, 2026.

Adaptation

Decision-directed identification under environment shift

Another line of my work asks what an agent actually needs to identify when the environment changes. The goal is to focus adaptation on information that matters for downstream decisions rather than treating every change in the environment as equally important.

What Must Be Learned to Adapt? Decision-Directed Identification Under Environment Shift. Under review, ICLR 2027.

Multimodal AI & vision

Lexical sensitivity in open-vocabulary detection

I am interested in how multimodal models respond to changes in language. In open-vocabulary object detection, this includes measuring how lexical choices in prompts affect model behavior and the reliability of reported detection performance.

LexiDet-OVD: Benchmarking Lexical Sensitivity in Open-Vocabulary Object Detection. Under review, WACV 2027.

Robotics & intelligent systems

AI-assisted irrigation and field deployment

During my M.S. and Graduate Student Researcher work in the UC Merced Robotics Lab, I contributed to OrchSmart, an intelligent irrigation-control system designed to reduce freshwater consumption in almond orchards. This work also formed the basis of my M.S. thesis and is associated with a pending UC patent case.

OrchSmart: Intelligent Control and Long-Term In-Field Test of Smart Irrigation in Orchards. Conditionally accepted, SenSys 2027.
Design and Implementation of an Automated System for AI-Agent based Optimized Irrigation in Almond Farms. M.S. thesis, 2025.
Patent pending: Systems and methods for Intelligent Irrigation Control, U.C. Case No. 2026-603 (2420-108 PRO).

Engineering background

Systems perspective

Before graduate school, I spent nine years building production software systems in healthcare and life sciences. That background continues to shape how I approach research: I value methods that are measurable, implementable, and connected to the behavior of real systems.

Research outputs

See the complete list of manuscripts, conference submissions, preprints, and thesis work.

Browse publications →