ECE Seminar: Udayan Das

ECE Seminar: Udayan Das

Title: Software Engineering in the Age of AI

Abstract: While we hear a lot about the impending demise of the Software Engineering profession and how Software Engineers are going to be replaced by AI, this talk provides a reality check. While there is extensive use of code generation and Agentic AI, there have also been considerable failures and security risks caused by premature deployment in some cases, and the lack of proper guardrails in others. Far from this being the moment for Software Engineers to recede into the sunset, this is the moment for the Engineering in Software Engineering to shine. Broad considerations of the appropriateness, test infrastructure, deployment boundaries, careful role-based access parameters, and CI/CD principles applied to contemporary Software development and deployment. Do not throw away the accumulated knowledge and understanding of the Software Engineering profession built over decades.

Bio: Udayan Das is an Associate Professor and the Founding Program Director of Computer Science at Saint Mary's College of California. He has a decade-and-a-half of teaching experience and 8.5 years of program management and administration experience. He loves teaching, creating new courses that address the needs of the day, and by extension rethinking the design of CS and CS-adjacent programs. His teaching methodology always attempts to center fun in learning while thinking about how to support students from diverse backgrounds to become successful in computing. He centers ethics and social justice in his teaching, and challenges students to not only consider harm reduction in the use of computing technology, but to go beyond and consider how they could be designing and developing systems that center values such as ethics and social justice right from the get go and throughout design, development through deployment. His research has been in Technical Language Processing, approaches to solving the factual inaccuracy and factual drift problems of LLM-GPTs, using knowledge-graph backed methods to ensure factual accuracy and micro-document/clause-level citability of source information. Extending AI to address difficult problems in healthcare and legal applications.