Multi-Modal Reinforcement Learning Algorithms for Improving Context-Sensitive Closed-Loop Blood Glucose Control for Type 1 Diabetics
My PhD project will focus on applying and developing state-of-the art reinforcement learning (RL) algorithms to improve blood glucose control for people with Type 1 Diabetes (T1D). Hybrid closed-loop insulin delivery systems have shown success in performing automated insulin dosing, but these devices are significantly limited in practical settings and require frequent interaction from users. My research will build on the current literature, to develop multi-modal algorithms capable of considering a diverse range of events in their decision-making, such as illness, activity, stress, and non-engagement. Within this topic I also hope to explore methods for achieving safe and sample-efficient policy learning and reducing user burden.
In 2020 I graduated from the University of Warwick with an integrated master’s degree in Physics. During my studies I was awarded the North American scholarship, entitling me to study Data science at Harvard University for half a semester. My master’s thesis focussed on designing and optimising a compact magnetometer, utilising an ensemble of nitrogen vacancy centres to achieve high levels of magnetic sensitivity. In addition, I worked with the engineering research group, WMG to evaluate the efficacy of Thixoforming for shaping steel. My research interests in Digital Health are primarily focussed on the use of machine learning for optimising clinical decision making. However, I have also participated in projects exploring the role of technology in maternal wellbeing and developing apps to allow clinicians to better manage heart failure in the community.