Human-centered AI
for health, wellbeing & work

We aim to develop wearable foundation models for personalized health monitoring using physiological and contextual data collected in everyday life.

EnPULSE (Enabling Platform for User Logging & Sensing Environment) is an open-source platform for mobile and wearable health research, which helps researchers run and monitor in-the-wild studies with less engineering overhead.

Digital Columbus uses multi-agent AI to interpret longitudinal patient-generated data and support psychiatric prognostic reasoning, preserving evidence and uncertainty for clinician review. AI Interviewer gathers psychiatric histories and screens for risks before visits.

We develop Physical AI for factory operations by integrating Factory-VLM perception, anomaly detection, and factory knowledge to assess risks, prioritize responses, and support safe human-robot collaboration through HMI and multi-agent coordination.

Meta-Scientist explores AI researcher clones for scholarly consultation, examining representation, reciprocity, and diversity. AI Science Hub investigates human digital twins as research participants, reviewing their construction, use, and validation.

We explore the mental model of older adults when interacting with a mobile agent, develop a language model specialized in understanding their utterances, and research on interactive methods for error recovery.

We explore agentic AI companions that sense daily context to deliver just-in-time adaptive interventions (JITAI) for well-being, while learning from user responses to remain personalized, safe, and non-intrusive.

We focus on a real-time mental health management system for emotional labor workers, such as call center agents. It uses sensors to monitor and assess their physiological and perceived stress states to provide personalized mental health management.