Abstract
The discipline of Software Engineering is being disrupted with the advent of large language models (LLMs) and agentic AI. There is a plethora of research, examples and testimonials out there on this. We will pull the curtain a bit more and look at the world of enterprise software – large application development and maintenance scenarios that are the primary interest for major enterprises and organizations across sectors and across the world. In this setting there are unique challenges that emerge related to data scarcity, complex application architecture, legacy programming languages, application modernization and working within the requirements imposed by the business. We will see examples of how LLMs and generative AI are being leveraged to address these challenges, but how it is important to combine these with algorithmic advances and program analysis to enable production use of such AI-based solutions.
Bio
Dr. Amith Singhee is the Director of IBM Research, India, and Chief Technology Officer for IBM India and South Asia. As Director, he sets the strategy for the Research division in IBM India for driving forward-looking innovation that fuels growth for IBM’s products and services. This includes foundational research in the areas of Hybrid Cloud, AI, Quantum Computing, Security, and Sustainability. As CTO, Amith engages with the regional ecosystem in academia and industry to represent IBM’s technology vision. On the technical front, Amith leads the research strategy worldwide for IBM Research in the area of AI for Software Engineering, and works across IBM’s offerings and clients to bring research innovations to bear in the market and real client scenarios.
Amith is an alumnus of IIT, Kharagpur, and has a Ph.D. in Electrical and Computer Engineering from Carnegie Mellon University, USA.
Abstract
Software development is a human activity and understanding software requires understanding humans that create it, both their individual experiences and the ways they work as part of a team. Of course, these two cannot be completely separated since the ways we feel and work as individuals influences and is influenced by the ways we collaborate and communicate. What makes this even more complicated is that nowadays teams tend to involve both human and artificial developers, namely bots. To illustrate these two sides of our research in this talk we start by discussing two recent studies focusing on individual experiences and teamwork, and then we reconsider these studies from a methodological perspective: we zoom in on how these studies have been conducted and what trade-offs design decisions have been embedded in these studies. Based on this reflection we sketch the space for future research on human aspects in software engineering.
Bio
Alexander Serebrenik is a full professor of social software engineering at the Eindhoven University of Technology, The Netherlands. His research goal is to facilitate evolution of software by taking into account social aspects of software development. His work tends to involve theories and methods both from within computer science (e.g., theory of socio-technical coordination; methods from natural language processing, machine learning) and from outside of computer science (e.g., organisational psychology). The underlying idea of his work is that of empiricism, i.e., that addressing software engineering challenges should be grounded in observation and experimentation and requires a combination of the social and the technical perspectives. Alexander has co-edited two books “Evolving Software Systems” (Springer Verlag, 2014), “Equity, Diversity, and Inclusion in Software Engineering: Best Practices and Insights” (APress, 2024) and co-authored more than 270 scientific papers and articles. He is actively involved in organisation of scientific conferences and is member of the editorial board of several journals. He has won multiple best paper and distinguished reviewer awards. Alexander is a senior member of IEEE and a member of ACM. Contact him at a.serebrenik@tue.nl.
Abstract
As AI agents become more and more prevalent in everyday technology, more and more individuals -- from every walk of life, at every level of education, across the entire socioeconomic spectrum, of every gender, race, ethnicity and age -- will need to make decisions about which agent(s) to use, when and how, and to what extent using them is the best path forward. The "mission" this talk explores is how we can enable such diverse individuals to make such decisions in ways that make their lives better instead of worse. For example, should I use an agent to enable me to be a remote caregiver for my grandmother, or should I move in with her? Should I buy semi-self-driving car X, or semi-self-driving car Y, or stay entirely manual? Will using one of these systems cost someone's life? Will it so destroy someone's privacy that their lives become filled with fear and harassment? Will my child become less intelligent over time if I give her access to LLM-powered "homework helpers"?
Bio
Margaret Burnett is a Distinguished Professor at Oregon State University. She began her career in industry, where she was the first woman software developer ever hired at Procter & Gamble Ivorydale. A few degrees and start-ups later, she joined academia, with a research focus on people who are engaged in some form of software development. She was the principal architect of the Forms/3 and FAR visual programming languages, and co-founded the area of end-user software engineering, which aims to improve software for computer users who are not trained in programming. Her end-user software engineering work included producing seminal work in actionably explaining AI to ordinary end users. She co-leads the team that created GenderMag, a software inspection process that uncovers gender inclusiveness issues in software from spreadsheets to programming environments. Her newest projects related to GenderMag include the InclusiveMag meta-method, SocioeconomicMag, and a new analytical approach to intersectionally inclusive software. Burnett is an ACM Fellow, a member of the ACM CHI Academy, and an award-winning mentor. She has served in over 50 conference organization and program committee roles. She was recently honored with the 2022 IEEE CS TCSE Distinguished Women in Science and Engineering (WISE) Leadership Award and the Grace Hopper Conference's ABIE Tech Leader Award.
Abstract
The problem of synthesizing specifications of undefined procedures has a broad range of applications. However, the usefulness of the generated specifications depends on their quality. In this talk, I will present our work on finding correct and maximal specifications using an approach called infer-check-weaken. This approach first infers a correct specification through a logical synthesis technique. A maximality check follows to determine whether the specification is the weakest. If not, the specification is weakened. We observe that our approach is effective across a range of benchmarks, including those with arrays.
Bio
Sumanth Prabhu works as a scientist in the Foundations of Computing group at TCS Research. He did his Ph.D. from the Indian Institute of Science, Bangalore, where he was advised by Deepak D'Souza. His current research focuses on formal verification and automated synthesis using constrained horn clauses.