We build NLP tools that measure mental health from language, so that monitoring no longer depends only on infrequent self-report.

We study what is needed to deploy large language models safely in mental health care, from real-world evaluations of conversational agents to frameworks for their development and regulation.

We examine how patients and clinicians interact with conversational AI: how people conceptualize it, how people use it, and where it fails.

We develop machine learning methods that go beyond diagnostic categories to describe how symptoms are organized and how they change over time.

We study telemedicine and other digital delivery methods that expand access to mental health treatment, using large real-world samples.