Important: It has come to our notice that some fraudsters are trying to misuse the names of IITH faculty members for defrauding educational Institutes and other organizations. Please beware! ×

Hybrid Intelligence for Software Engineering: Integrating Classical Tools with Generative/Agentic AI

Title of the Talk: Hybrid Intelligence for Software Engineering: Integrating Classical Tools with Generative/Agentic AI
Host Faculty: Dr. M V Panduranga Rao
Speaker: Atul Kumar
Date: 17 Jun 2026
Time: 11:00 am
Venue CSE LH1

Abstract
The emergence of frontier large language models (LLMs) and agentic AI systems has transformed the landscape of software engineering. Tasks that have traditionally been challenging for classical techniques - such as large-scale code transformation, migration, and maintenance - can now often be addressed effectively using modern generative AI. However, reliability, predictability, and controllability remain significant challenges. In practice, improving performance frequently relies on prompt engineering and iterative experimentation, offering limited guarantees about correctness and consistency.

At the same time, enterprises possess a rich ecosystem of software engineering tools and techniques developed over decades, including program analysis, static and dynamic analysis, testing frameworks, and domain-specific rule-based systems. While these approaches may not solve every problem end-to-end, they provide valuable structure, precision, and trustworthiness.

Our research explores how to combine classical software engineering methods with modern agentic and generative AI systems to leverage the strengths of both paradigms. Rather than viewing AI and traditional techniques as competing approaches, we investigate hybrid architectures in which program analysis and other established methods guide, constrain, validate, and enhance AI-driven workflows.

In this talk, I will present several real-world enterprise-scale problems that we have solved - or are actively addressing - using this hybrid approach. These case studies illustrate how the integration of classical software engineering techniques with agentic AI can improve reliability, scalability, and practical impact in industrial settings.

Bio

Atul Kumar is a Senior Research Scientist and Research Group Manager at IBM Research, where he leads research in AI-assisted test automation, code understanding/explanation, code transformation and other areas in AI for Code domain. His primary research interests span Artificial Intelligence, Software Engineering, Internet Technologies and Distributed Systems.

With over 23 years of experience, Atul has worked with leading research organizations including IBM Research, Siemens Corporate Technology, Microsoft R&D, Accenture Technology Labs, and ABB Corporate Research. He has authored over fifty publications in top-tier computer science and automation journals and conferences, and holds more than two dozen patents (issued or filed).

He currently serves as the Chair of ISOFT (ACM SIGSOFT India Chapter) and is a member of the Executive Committee of IKDD (ACM SIGKDD India Chapter).

Atul received his MTech and Ph.D. degrees in Computer Science and Engineering from IIT Kanpur in 1998 and 2003, respectively. He is a Senior Member of both IEEE and ACM.