General Profile:

I have worked for 15 years in industry as a founder, software engineer, technical leader, and advisor to engineering teams. During that time, I lead teams to deliver innovative software solutions across a variety of domains, technologies, languages, and platforms using agile and lean methodologies. Before coming back to academia, I spent 6 years building my own medical device startup with the vision to help people decide how much insulin to take and when. To this day I continue to mentor female engineers at the start of their careers.

Research Project Summary:

The aim of my PhD is to use AI to help expand upon what is known about insulin needs in people with Type 1 Diabetes (T1D) and the factors that impact it. We aim to combine expert knowledge about what contributes to an interesting pattern with machine learning methods capable of understanding such patterns. With our research, we hope to be able to uncover opportunities for new medical research into improving the treatment of T1D.

Time series are ubiquitous in health data, but gaps remain in current machine learning methods to be able to use them to extract actionable insights about human health. Challenges faced include missing data, irregular readings, high dimensionality, shifting distributions over time and the need for transparency of the algorithm’s decision-making for medical experts to trust the machine’s advice.

Type 1 Diabetes is a chronic condition where the body produces little or no insulin, a hormone required for the cells to use blood glucose for energy and to regulate BG levels in the body. Finding the correct insulin dose and time remains a complex, challenging and yet unsolved control task.

To be able to achieve the aim of the PhD we will need to overcome current limitations in methods that can learn actionable insights from time series data. These include the ability to derive patterns from the irregularity of the data, understand patterns in time beyond sequentially and the ability to present these findings in a human-interpretable way.