Wearable-powered mental health companion for Nepali students and professionals

Sahara (साहारा) tracks physiological signals from wearables — HRV, sleep, resting heart rate — and reaches out when the data suggests something is off, often before the user notices it themselves.

In communities where mental health stigma is high, people tend to wait until things are bad before asking for help. The goal was to intervene early and quietly, in a way that made cultural sense for Nepali students and professionals.

Trained a LightGBM classifier on 71-user Fitbit Sense data using 30+ physiological features to predict when someone might need a check-in. Two AI agents, one per cultural context, run on GPT-4o/Claude with a RAG pipeline over counseling guidelines. One tap from the notification to the right agent.