Well mind: AI-powered mental health monitoring and prediction

Psychological illnesses such as depression, anxiety, and stress are still prevalent but difficult to track objectively. That&s;s why we created Well Mind, a computer program that forecasts mental well-being with clinical-grade precision in a multi-modal fashion. It combines full-body kinematic observations with Kinect V2, language use from social media through DeBERTa encoding, and user-reported behavioral information e.g. sleep, diet, and suicidal thoughts. Kinematic paths are forecast using a gradient boosting model, and questionnaire answers are label-encoded and normalized for modeling. Forecasts from all data sources are combined with a weighted majority voting strategy. The system is implemented through a FastAPI REST interface for seamless integration into clinical practice. When applied to a held-out cohort, Well Mind was 85.3% accurate, 84.7% precise, 85.1% recall, and ROC AUC of 0.89—meaning that it did have promise to be an excellent instrument for monitoring of mental health.