The proposed research project focuses on multi-sensor such as heart rate, body temperature, oxygen saturation level, and inertial sensor data-based sleep stage classification, and tremor detection in real-time for preventive health devices. The goal of the proposed research is to build robust and energy-efficient machine learning and deep learning-based approaches that extract and analyze the significant information from the multisensor data coming from wrist-based health devices to help Parkinson's patients with tremors and an individual with sleep quality tracking.