The exponential rise in data has spurred significant interest in developing personalized machine models for a seemingly infinite range of applications. However, in the case of safety critical systems such as autonomous medical systems, classical algorithms have failed due to complications spanning from limited data to the requirement that systems be explainable and verifiably safe. In this talk, I will present an overview of our recent contributions in how formal verification can be coupled with modeling and controls to broadly answer “How can we build useful, personalized autonomous medical systems which can act safely in the presence of uncertainty?”. While our work focuses on a range of applications, this talk is rooted in the fascinating example of artificial pancreas systems, a set of interconnected devices which seek to automate insulin delivery for individuals with type-1 diabetes. Advances in this field have the potential to significantly improve health outcomes and reduce individual mental burdens, and applications of the methods developed extend beyond the clinical setting to systems such as autonomous aircraft and energy grids.
Dr. Taisa Kushner is an NSF/CRA Computing Innovation (CI) Fellow at Oregon Health & Science University, and Adjunct Assistant Professor at Portland State University. Her research lies at the intersection of formal verification, machine learning and cyber-physical systems control, focusing broadly on applications in autonomous medical devices. She received her PhD in Computer Science and Interdisciplinary Quantitative Biology from the University of Colorado Boulder, under the advising of Dr. Sriram Sankaranarayanan. She also holds an MSc in Applied Mathematics from the University of Colorado Boulder, and a BA in Mathematics and Art from St. Olaf College. Prior to her current role, Dr. Kushner held research appointments at a range on institutions including the University of Minnesota, University of Virginia Center for Diabetes Technology, and Microsoft Research.
Galois was pleased to host this tech talk via live-stream for the public. A video of the presentation can be found above.