Zenly

Zenly is an AI-powered mental health application designed to help people understand how their daily habits may influence their stress levels. Many people track sleep, health, work, or lifestyle habits, but still struggle to identify which patterns contribute to stressful periods. Even when stress is recognized, it can be difficult to know what practical action to take.

Zenly addresses this problem through two connected AI components. First, a machine-learning model analyzes user information and predicts the number or category of stressful days a person may experience. The model was developed using a large health dataset from the American Centers for Disease Control and Prevention, containing approximately 450,000 respondents and hundreds of health-related factors. Because the target data contained a high number of zero-stress responses and an uneven distribution, the team explored oversampling and a combined classification-to-regression approach.

The final CatBoost-based model achieved 81% classification accuracy and an RMSE of 5.3, improving substantially compared with the baseline model.

The second component is an LLM-powered chatbot that turns predictions into understandable and actionable guidance. The chatbot uses expert mental-health documents through a retrieval-augmented generation workflow and was also compared with a fine-tuned model. While the RAG version produced broader and more contextual answers, the fine-tuned model delivered shorter, more practical recommendations.

Together, these components allow Zenly to move beyond simple stress tracking. The product helps users recognize patterns, anticipate difficult periods, and receive personalized guidance for improving their daily balance.