Rentrée SERIQ

SERIQ's rentrée took place on October 29, 2026, and brought together about 100 people. We warmly thank everyone who attended. Photos will follow.

SERIQ expands the SEMTL research network to work on software engineering for the digital society. For its rentrée, we invited researchers, students, industry partners and institutional representatives to guest talks from research and industry, followed by a panel on software research for society. Benoit Baudry then presented the history and scope of SERIQ, followed by remarks, before a networking reception with student posters.

Programme

1. 3:45–4:00 p.m.

   Registration

   Check-in for registered attendees.

2. 4:00 p.m.

   Opening

   Welcome and introduction to the afternoon.

3. 4:15–5:30 p.m.

   Guest talks

   Four speakers from research and industry each present their view on software research in 15 minutes.

   • 

     Daniel Lemire

     Université TÉLUQ

     Open source as a research method: from the scientific paper to billions of devices

     Abstract

     Open-source libraries such as simdjson and simdutf now run in Node.js, in browsers and in databases, serving billions of people every day. The linear model of innovation — the professor publishes, the engineer reads, the customer consumes — rarely describes what actually happens. As Michael Stonebraker pointed out, ideas come mostly from contact with real users and their concrete problems. Open source is not merely a way to disseminate results: it is a research method in its own right. It enforces reproducibility, since the code and the benchmarks are public and anyone can check them. Code reviews, bug reports and pull requests amount to a peer review that is often more demanding than many review committees. It also brings in real-world data: adversarial cases, exotic architectures and unforeseen workloads. The story of simdjson reverses the expected order: first the code, a blog post, GitHub and a few posts on X, then industrial adoption, and only then the scientific paper. The cycle repeats: use reveals the next bottleneck, hence the Eisel-Lemire algorithm for floating-point numbers and simdutf for UTF-8. The approach has its tensions, however: maintenance is thankless and never-ending, and academic incentives count papers, not concrete contributions.

     Biography

     Daniel Lemire is a professor of computer science at the Université du Québec (TÉLUQ). His research and engineering work covers high-performance computing, efficient data structures and software performance optimization. He is the author of widely adopted open-source libraries: his algorithms and his code are found in several standard libraries and major systems, including .NET, Rust, Go, Node.js, Google Chrome and Safari. He is among the top 2 % of most-cited scientists worldwide according to the Stanford University / Elsevier ranking. He is an editor of the journal Software: Practice and Experience (Wiley, founded in 1971) and received the Université du Québec’s 2020 Research Excellence Award for his work on accelerating JSON document processing. On GitHub, he is among the 1000 most-followed developers out of more than 100 million. He is @lemire on X and writes a weekly blog at https://lemire.me/.

   • 

     Lionel Briand

     University of Ottawa

     AI Is Breaking Your Software Testing Strategy, and It Can Also Fix It

     Abstract

     Organizations are racing to embed AI into their products, from LLM-powered assistants to autonomous agents and machine learning features. But AI-based systems don’t behave like traditional software. The same input can produce different outputs, “correct” is often a matter of judgment, and failures can surface as subtle quality problems, bias, or hallucinations rather than obvious crashes. Many teams find that their existing testing practices, metrics, and release gates no longer give them the confidence they need to ship. For leadership, this translates into real business risk: reputational damage, compliance exposure, and slower time to market.

     AI also offers a way forward. Used well, it can automate much of the work that makes quality assurance slow and expensive: generating realistic test scenarios and data, evaluating open-ended outputs at scale, continuously probing systems for weaknesses, and helping teams pinpoint the root causes of failures faster. The result can be broader coverage, faster feedback, and lower QA costs, for AI-based systems and conventional software alike.

     This talk offers a clear-eyed look at both sides. For executives, it frames the new quality risks AI introduces and what it takes to manage them, including where investment pays off and where human oversight remains essential. For practitioners, it presents concrete techniques and tooling patterns for testing AI-based systems and for using AI to strengthen existing QA workflows, along with the pitfalls to avoid, such as over-trusting AI evaluators.

     Biography

     Lionel C. Briand is a professor of software engineering and holds shared appointments at (1) The University of Ottawa, Canada, and (2) The Lero Centre—the Research Ireland center for software research—hosted by the University of Limerick. He holds a Distinguished Research Chair (Trustworthy AI-enabled Software Systems) and directs Lero. Lionel Briand was elevated to the ranks of IEEE Fellow and ACM Fellow for his contributions to software testing and verification. He received the IEEE Computer Society Harlan Mills Award, the ACM SIGSOFT Outstanding Research Award, and the IEEE Reliability Society Engineer of the Year Award. Additionally, he was awarded an ERC Advanced Grant in 2016 for modeling and testing cyber-physical systems—the most prestigious individual research award in the European Union—and was elected a fellow of the Academy of Science, Royal Society of Canada, in 2023, Academia Europaea—the European Academy of Science, in 2025, and the inaugural ACM SIGSOFT Software Engineering Academy in 2026.

   • 

     Marsha Chechik

     University of Toronto

     Engineering the Digital Society: From Human Intent to Dependable Systems

     Abstract

     Digital systems shape access to services, safety, privacy, and opportunity. Their failures are often described as bugs, but many begin earlier: with assumptions left unstated, stakeholders whose needs were not represented, or components that do not work together as intended. Software engineering gives us ways to make those assumptions explicit, translate human and societal needs into system requirements, and gather evidence that systems meet them as they change. This challenge becomes more urgent as AI helps produce software at increasing speed. I will draw on examples from safety-critical systems and my work on social, legal, ethical, empathetic, and cultural requirements to argue that dependable software depends on more than correct code. It depends on sustained alignment between systems and the people who rely on them. SERIQ will help build that capacity through research, collaboration, and the training of people who can put it into practice.

     Biography

     Marsha Chechik is a Professor in the Department of Computer Science and Bell University Chair in Software Engineering at the University of Toronto. She received her Ph.D from the University of Maryland in 1996.

     She has authored over 200 papers in formal methods, software specification and verification, computer safety and security and requirements engineering. Her current interests are in software assurance and reliability in a variety of domains including automotive, robotics and healthcare.

     Professor Chechik has been Program Committee Co-Chair of many top international conferences in software engineering, verification and formal methods, including FM’23, ESEC/FSE’21, ICSE’18, VSTTE’16, TACAS’16, ASE’14, FASE’08, CONCUR’08. She also serves as Associate Editor in Chief of International Journal on Software and Systems Modeling and IEEE Transactions on Software Engineering.

     Professor Chechik is Fellow of Automated Software Engineering, Distinguished Member of ACM, Member of IFIP 2.3 Working Group on Requirements Engineering and Vice Chair of ACM Special Interest Group on Software Engineering (SIGSOFT).

   • 

     Yves Daoust

     Ferme d'Hiver

     From punch cards to AI agents: How have the organization, costs, and business models of software products evolved over the past 40 years, and why is it different this time?

     Abstract

     This talk will retrace a historic shift in the balance between hardware and labor costs, the transformation of development teams, and the rise of the web, agility, and the cloud as business models. It will then place this shift toward AI-assisted support within a broader trend. Based on recent data, I will distinguish between the creation of new products and the maintenance of code that has already been mastered, discuss the operating costs of AI agents, both in the cloud and on-premises, and conclude by noting the emergence of industrial markets that were long considered too complex and underdeveloped, but are now undergoing a major transformation.

     Biography

     Yves Daoust is an engineer who worked in telecommunications before turning to agriculture. He founded Ferme d’Hiver in 2018 and has operated it since.

4. 5:30–6:00 p.m.

   Break

   A short break before the panel.

5. 6:00–6:45 p.m.

   Panel

   The speakers discuss Software research for society: needs and opportunities.

   • Daniel Lemire
   • Lionel Briand
   • Marsha Chechik
   • Yves Daoust
   • Martin Robillard Moderator

6. 6:45–7:15 p.m.

   Presentation of SERIQ

   Benoit Baudry, director of SERIQ, presents its history and scope. Representatives of the universities then give remarks:

   • Frédéric Bouchard, Dean of the Faculty of Arts and Science, Université de Montréal
   • Daria Camilla Boffito, Associate Director, Research and Research Support, Polytechnique Montréal
   • Brigitte Pientka, Director of the School of Computer Science, McGill University
   • Alan Carter, Dean of Research, ÉTS

7. 7:15–8:00 p.m.

   Networking reception

   With around 35 student research posters.