Modern computing systems that interact with patients and clinicians, commonly referred to as medical cyber-physical systems (MCPS), are safety-critical embedded systems that feature tight coupling between communication and computation used to control complex, dynamic, and uncertain physical/physiological plants. Learning-enabled MCPS additionally incorporate components whose behavior is driven by “background knowledge” acquired and updated through a “learning process”. While empirical medical data is often a significant source of this background knowledge, it can also be limited, sparse, or “thin” due to small sample sizes, dataset shifts, anomalies, inter/intra-patient variability, and a limited understanding of the data generation process itself. Consequently, providing safety guarantees and predictable performance for learning-enabled MCPS in the presence of thin data is challenging. In this talk, I will present some of my recent work on techniques and tools for the design and analysis of safe learning-enabled MCPS with thin data. Specifically, in the context of learning enabled medical systems, I will present our Verisig tool for formal verification of closed-loop learning-enabled systems. Real-world case study evaluations and implementations covering control of type I diabetes, prediction of hypoxia in infants, and alarm suppression in intensive care units, illustrate the utility of my group’s recent work and give light to future research challenges.