Leveraging Formal Methods for Trustworthy AI-assisted Software Development

This is a hybrid in-person/remote event. Zoom link: https://umich.zoom.us/j/98689964031

Abstract: This talk will cover my academic research path from asking deep questions around our understanding of software developer’s cognitive load during program comprehension, to using proof-oriented programming to improve trust in AI-assisted programming tools, and how it has positioned me to creatively work on research problems using cross disciplinary approaches. I’ll talk about the current state of AI for code research, open problems around building trusted and auditable AI-software systems, and how formal verification and human-centered research techniques will be critical to building safe and reliable software systems of the future.

Bio: Sarah Fakhoury is a Postdoctoral Researcher in the RiSE group at Microsoft Research, Redmond. Her research centers around ways to improve the design and evaluation of software development tools to optimize the cognitive effort of programmers. She is currently interested in leveraging formal methods as a mechanism to provide end-user trust and verifiability of AI-assisted programming tools. Her approach to research is largely interdisciplinary, fusing: formal methods, empirical software engineering, machine learning, and human-computer interaction.