Abstract: Generative AI and Large Language Models (LLMs) are rapidly transforming software engineering, enabling AI agents to assist with increasingly complex development tasks. Yet a fundamental challenge remains: how can we trust AI systems that are known to hallucinate, make reasoning errors, and produce plausible—but incorrect—solutions? This talk presents our research on building trustworthy agentic AI by combining the creativity of LLMs with the rigor of static and dynamic program analysis, formal constraints, and meaningful human oversight. Rather than replacing software engineers, these hybrid human-AI systems enable safer, more transparent, and more reliable automation of complex software engineering tasks.
I will present several AI systems developed by our group for automated refactoring, software modernization, and large-scale code transformations, along with empirical results demonstrating substantial improvements over existing approaches and successful adoption by major open-source projects. I will conclude by sharing a broader vision for trustworthy agentic AI and discuss opportunities for interdisciplinary collaboration as we work toward establishing a large-scale Center on Trustworthy AI.
Bio: Danny Dig is an Associate Professor of Computer Science at the University of Colorado Boulder. Following an entrepreneurial leave at JetBrains Research, he leads research on trustworthy agentic AI and Generative AI for Software Engineering, combining large language models with program analysis and human oversight to build more reliable AI systems. Danny is the founder and Executive Director of the NSF Industry–University Cooperative Research Center on Pervasive Personalized Intelligence (PPI Center) and is leading a university-wide initiative to establish a large-scale Center on Trustworthy AI. His research has received 14 distinguished paper and impact awards, and software developed by his research group has been incorporated into widely used developer tools used by millions of software engineers every day.
