Towards Code-Aware AI
Abstract:
AI for Code has rapidly emerged as a powerful tool to enhance developer productivity, enabling breakthroughs in tasks like code generation and program repair. However, while these models show promise, ensuring software quality, robustness, and security remains a major challenge, and is often not the primary focus of current systems.
In this talk, I will discuss how neurosymbolic methods, which combine the strengths of machine learning with formal reasoning and program analysis, can address these gaps. I’ll present a class of AI agents that leverage both static and dynamic properties of code—such as control/data flow and runtime traces—to reason more effectively about program behavior. These agents enable deeper automation of critical tasks like debugging and program repair, offering a path toward trustworthy, explainable, and collaborative developer tools. I will illustrate these ideas through examples from both academic research and real-world deployments.
Bio:
Baishakhi Ray is an Associate Professor in the Department of Computer Science at Columbia University in New York, USA. Her research lies at the intersection of Artificial Intelligence and Software Engineering, with a particular focus on developing intelligent, trustworthy tools for improving software reliability, security, and developer productivity.
Dr. Ray’s work has been recognized with several honors, including the IEEE TCSE Rising Star Award, NSF CAREER Award, IBM, and VMWare Faculty Awards. Her publications have received multiple Distinguished Paper Awards at top venues, the ICSME Test of Time Award, and have been featured in CACM Research Highlights, reflecting the broad impact of her research on both the academic and industrial communities. She also serves as an Amazon Scholar, collaborating with the Amazon Q Agentic AI team to build intelligent developer agents grounded in real-world software engineering practice.