Researchers have explored developing AI systems that repurpose data from everyday devices to monitor symptoms of mental health disorders. However, the heterogeneous presentation of these symptoms across individuals presents challenges towards developing accurate symptom monitoring systems. Specifically, how can we develop systems that accurately detect patient-specific symptoms, and ensure reliable symptom monitoring? How can these systems support patients and their clinicians? In this dissertation, I present three studies focused on designing, developing, and evaluating AI-driven symptom monitoring technologies to address these challenges. In particular, this work centers on developing passive sensing AI systems. Passive sensing AI systems process the behavioral and physiological data passively generated and collected from everyday devices to estimate symptoms of mental health disorders. I close by discussing future work building on this research to develop data-driven technologies that support a more responsible and patient-centered healthcare technology development.