In a hospital large amounts of patient data are generated (e.g. medical images, test results, consultations) in a variety of form, volume and velocity. High impact decisions are made daily based on these patient data by healthcare professionals, who unavoidably are biased by personal experiences and opinions. Data science can help clinicians in this process by providing real-time clinical decision support systems. In this talk I will walk the audience through our experience in developing a practical data science pipeline at the NKI radiotherapy department. In particular, I will present the case of an anomaly detection system as a real-time ‘red-flag’ application for CT-based image delineations and how this complements the clinical workflow and supports the clinician in making more informed decisions. I will start with an overview of the relevant healthcare professionals in a radiotherapy department, including the ‘stakeholders’ and the ‘decision-makers’. I will then focus on how a data science project can be implemented, given the available resources in a hospital (e.g. people, databases, servers), as well as the restrictions often posed (e.g. data privacy laws, data silos). In addition I will spend some time discussing the veracity of clinical data and I will highlight some interesting cases. Lastly, I will present elements of the libraries, models and data pipelines deployed on this project and clinical insights that were revealed by our anomaly detection system.