Keynotes
Jordi Cabot - Vibe-Driven Engineering: when low-code fell in love with AI
Abstract: Low-code and model-driven engineering have long raised the level of abstraction in web development, improving productivity, quality, and maintainability of the software process. Today, the rise of Large Language Models (LLMs) and vibe coding promise an even faster way to web development: from English to code but with no guarantees on the scalability, security, robustness, … of the solution. In this keynote, I will argue that AI is not the end of low-code, but a valuable partner. Drawing on our experience with BESSER, our open-source low-code platform for hybrid systems development, I will discuss how AI and low-code can work together by using software models as the central bridge between humans, agents and reliable software outcomes.
Bio: Jordi Cabot is an FNR Pearl Chair and the head of the Software Engineering RDI group at the Luxembourg Institute of Science and Technology. He is also an Affiliate Professor in Computer Science at the University of Luxembourg. He has previously been at Inria, University of Toronto and the Open University of Catalonia among others. His core research goal is helping organizations build better software faster. This includes topics such as software modeling and low-code technologies, pragmatic formal verification, analysis of open source/open data communities and the role AI can play in software development (and vice versa). For more information, visit https://jordicabot.com
Senjuti Basu Roy - Rethinking Web Retrieval: From Items to Sets in the Age of Large Language Models
Abstract: Modern web applications increasingly leverage multi-modal data, including text, images, and structured attributes, to support complex user queries. In many scenarios, users seek coherent bundles of items rather than individual results, and the scoring criteria guiding these selections may be user-defined and expressed in natural language. As a result, retrieval systems must interpret user intent and evaluate candidate item sets holistically. This talk revisits the problem of Top-k set retrieval in such settings, where the goal is to identify the best set of k items that jointly satisfies user intent. Unlike traditional retrieval systems that rank items independently, composite retrieval requires set-level reasoning to assess compatibility, complementarity, and constraints across items.
I will present how large language models (LLMs) create new opportunities for enabling flexible semantic scoring over multi-modal data. However, integrating LLMs into retrieval pipelines introduces important challenges, including latency, monetary cost, and uncertainty in model outputs. In this talk, I discuss emerging directions for addressing these challenges, including cost-aware retrieval frameworks, decomposable scoring models that improve interpretability, and interactive pipelines that incorporate human feedback, pointing toward reasoning-augmented web systems for next-generation intelligent applications.
Bio: Senjuti Basu Roy is the Panasonic Chair in Sustainability and an Associate Professor in
the Department of Computer Science at the New Jersey Institute of Technology. Her research lies at the intersection of data management, information retrieval, and AI, with a focus on enabling large-scale human-machine analytics. Senjuti has published more than 90 research papers in leading conferences and journals in data management and data mining. She has served as Program Co-Chair of the ACM International Conference on Information and Knowledge Management (CIKM 2025), the International Conference on Extending Database Technology (EDBT 2027), and Co-Chair of the IEEE International Conference on Data Mining (ICDM 2025) Women in Science Forum. She is a recipient of the NSF CAREER Award and was selected as one of the 100 invited early career engineers by the U.S. National Academy of Engineering (NAE) in 2021.
Efthymia Tsamoura - Scalable Rule Mining Under Formal Guarantees
Abstract: Given a relational database, the problem of rule mining asks us to construct a set of rules that explain the facts stored in that database. This task is central to data management, knowledge graphs, responsible AI, and machine learning, with applications ranging from recommendation systems and knowledge graph completion to computer vision and natural language processing. Despite extensive research, state-of-the-art rule mining methods suffer from two critical limitations that hinder real-world adoption: scalability and the lack of formal guarantees. In practice, most techniques do not scale beyond a few thousand facts—even with specialized hardware. More importantly, they provide no guarantees to users regarding the quality of the mined rules.
This talk presents the first technique capable of mining rules from databases with millions of facts in seconds, using only commodity CPUs. Relying on a new result showing that graph patterns can be mined under \epsilon-guarantees in polynomial time in the number of graph nodes, the proposed technique offers users the ability to control the quality of the mined rules. The talk concludes with an extensive empirical comparison against state-of-the-art methods, showing that the proposed technique mines more accurate rules on CPUs within 1% of the runtime required by competing GPU-based approaches.
Bio: Efthymia Tsamoura is a Technical Expert at Huawei Labs. From 2019 to 2025, she was a Senior Researcher at Samsung AI, Cambridge, UK. In 2016, she was awarded a prestigious early-career fellowship from the Alan Turing Institute, UK, for her work on logic and databases, and before that, she was a Postdoctoral Researcher in the Department of Computer Science of the University of Oxford. Her main research interests lie in the areas of logic, knowledge representation and reasoning, and neurosymbolic learning, while her recent outcomes involve scaling symbolic reasoning to billions of triples, as well as addressing open problems in neurosymbolic learning. Her research has been published in top-tier AI and database venues (NeurIPS, ICML, SIGMOD, VLDB, PODS, AAAI, IJCAI, etc.). In 2024, Efi was invited by the Royal Society, UK, to the Frontiers of Science on AI meeting to discuss the risks of AI and ways to address them. More details can be found at https://tsamoura.github.io