We are transforming the field of mental health care through pioneering advancements in computational psychiatry, precision medicine, and digital health technologies.
Our research merges cutting-edge machine learning algorithms, digital phenotyping techniques, and clinical predictive analytics to create personalized and scalable solutions tailored to patient needs.
By leveraging routinely collectable data combined with advanced deep learning methods, we identify multi-modal biomarkers and risk factors that guide stratified treatments and preventative strategies.
Our focus on accessibility, equity, and practical implementation positions our innovations at the forefront of improving clinical decision-making, fostering global collaborations, and shaping the future of personalized mental health care.
Digital Biomarkers Extraction Toolkit
Our in-house Digital Biomarker Extraction Toolkit converts raw, pre-consented interview recordings into a unified stream of behavioral signals spanning face and head dynamics, gaze and oculomotor cues, fine movement, speech acoustics, and language patterns.
The pipeline standardizes inputs, optionally performs speaker diarization, and runs parallel video, audio, and NLP analytics to produce both time-series and session-level summaries designed for statistical modeling and machine learning.
Built for clinical and behavioral research, it emphasizes robustness and seamless integration with downstream analysis while scaling from single sessions to large cohorts.
Real-time PTSD risk from routine triage text
A clinically informed, privacy-preserving AI system that converts a patient’s triage narrative into an immediate risk estimate for post-traumatic stress disorder, along with a concise, human-readable rationale.
It runs on open-source models within a HIPAA-compliant environment, requires no additional questionnaires or staff time, and integrates into existing ED workflows without changing how clinicians document care
Our framework is designed for deployment at scale: it ingests routine free-text, supports batch or streaming inference, and exports structured results to downstream registries, analytics, or care-pathway tools. The result is a fast, interpretable assessment that helps teams identify at-risk patients during the narrow window when early intervention can change outcomes.