Cancers are complex, continuously evolving diseases unique to each individual patient. Yet, most drugs are administered according to fixed, “one-size-fits-all” strategies that always aim to deliver the maximum tolerated dose until it fails due to toxicity or resistance. “Adaptive Cancer Therapy” (AT) is a novel approach which seeks to delay drug resistance by personalising when and how much drug is given, based on the tumour’s response dynamics. After initial success in prostate cancer, there are now important questions about which patients would benefit from AT, and how should we optimally adapt therapy? In this 2-year fellowship, I will develop a computational framework to: i) understand and track resistance evolution from a patient’s tumour burden data (volume, blood biomarkers, ctDNA), and ii) translate this knowledge into personalised dosing strategies to slow resistance evolution and improve quality-of-life. To predict whether a patient will benefit from AT we need a firm, quantitative understanding of the dynamics of resistance evolution. Under supervision of Prof Trevor Graham at the ICR, I will develop a software package (B-REDi) which integrates a family of mathematical models and Bayesian inference to learn about the evolutionary route to resistance in a patient. Subsequently, I will deploy this package to investigate how AT is altering resistance evolution in a clinical trial in ovarian cancer (ACTOv). Finally, I will use deep reinforcement learning to explore how our calibrated models can guide treatment decisions even under uncertainty about resistance mechanisms. In the future, I plan to establish my own lab to develop computational tools to mitigate resistance and toxicity through schedule personalisation. This project will serve me as a steppingstone towards this goal, and it will contribute novel tools and insights towards a future in which mathematical models are used akin to weather forecasts to inform clinical decision-making.