51²è¹Ý

Zoom Link: https://zoom.us/j/94636927417?pwd=korRkwpSqXtvPM1agbO98jQOFUZbGX.1

Abstract: Code assistants powered by large language models have rapidly gained popularity, demonstrating impressive capability in generating small, well-scoped, greenfield code. However, their effectiveness drops sharply when applied to large, long-lived enterprise applications, where scale, legacy design patterns, and deep business logic pose significant challenges.

In this talk, we share insights from our experience building and evaluating IBM watsonx Code Assistant for Z on real-world, large-scale mainframe applications. We discuss the technical and architectural obstacles that arise in brownfield environments—such as program size, cross-program dependencies, legacy constructs, and domain-specific frameworks—and outline emerging opportunities and directions for making code assistants truly effective for enterprise modernization.

About the Speaker: Yogish Sabharwal is a Distinguished Engineer at IBM Research India, where he spearheads the AI for Code team. His leadership was pivotal in driving the core innovations behind the IBM watsonx Code Assistant for Z. With a career spanning over 25 years, Dr. Sabharwal has made significant contributions across the domains of Artificial Intelligence, High-Performance Computing, and networking.

He holds a Ph. D. from the Indian Institute of Technology (IIT) Delhi, specializing in approximation algorithms for computational geometry. Beyond his corporate leadership, he is a dedicated educator, and has served as adjunct faculty at IIT Delhi and Delhi University.

We look forward to your active participation.