Mobile health is an emerging field that allows for real-time monitoring of individuals between routine clinical visits. Among others it makes it possible to remotely gather health signals, track disease progression and provide just-in-time interventions. Consumer grade wearable sensors can remotely gather health signals and other time series data. While wearable sensors can be readily deployed on individuals, there are significant challenges in converting raw sensor data into actionable insights. In this dissertation, we develop machine learning methods and models for personalized health monitoring using wearables. Specifically, we address three challenges that arise in these settings. First, data gathered from wearable sensors is noisy making it challenging to extract relevant but nuanced features. We develop probabilistic graphical models to effectively encode domain knowledge when extracting features from noisy wearable sensor data. Second, prediction models developed on one population in lab settings may not generalize to other populations in field settings. We develop domain adaptation techniques to improve lab-to-field generalizability. Third, collecting ground truth labels for health monitoring applications is expensive and burdensome. We develop active learning methods to minimize the effort involved in collecting ground truth labels. We evaluate these methods and models on two case studies: cocaine use detection and human activity recognition.