Selected work spanning wearable sensing, ecological momentary assessment, personalized machine learning, and digital phenotyping.
A longitudinal digital health study combining Fitbit sensing, ambulatory blood pressure measurements, and ecological momentary assessment to model person-specific stress and blood pressure reactivity.
Methods and infrastructure for using smartphone and wearable data to characterize clinically meaningful changes in depression, anxiety, stress, and related outcomes over time.
Personalized and self-supervised models that combine wearable biosignals with EMA labels to study real-time detection of craving and substance use under limited labeled data.
Person-specific machine learning and multimodal approaches for modeling affective signals from video and physiological data.