I build intelligent systems that understand human behavior beyond the lab.
I’m a Computer Science Ph.D. student at Georgia Tech, advised by Prof. Thomas Plötz. My work combines generative AI, embodied agents, and multimodal sensing to create scalable, privacy-conscious tools for understanding activity, engagement, and behavior.
Georgia Tech
From simulated worlds to real human moments
I develop methods that make human-centered sensing more capable and more practical: generating virtual sensor data when real data is scarce, learning from physiological and behavioral signals, and designing systems that work across people, environments, and modalities.
Generative sensing
LLMs and embodied agents generate diverse, privacy-preserving motion and ambient sensor data for human activity recognition.
Multimodal understanding
Wearable, physiological, visual, and interaction signals reveal engagement and behavior in realistic settings.
Inclusive human-centered AI
Data and models designed around real people—including accessibility and health contexts—rather than idealized benchmarks.
Current projects
EduGage
Estimating momentary engagement during self-guided video learning using synchronized wearable, physiological, eye-tracking, and behavioral signals.
Read the paper
AgentSense
LLM-guided agents live out diverse routines in simulated smart homes, producing scalable ambient sensor data without collecting it from people.
Read the paper
IMUGPT 2.0
Turning natural-language activity descriptions into diverse virtual IMU signals through motion synthesis and cross-modality transfer.
Read the paper Code