| [9:00 - 09:15] Opening Session |
|---|
| 9:00 | Welcome
The CompSys 2026 organizers |
| [09:15 - 10:15] Session 4: Sustainable AI and Digital Twin Systems (Chair: Anuj Pathania) |
|---|
| 09:15 | Enabling Fine-Grain DVFS for Multi-Kernel GPU Workloads
Jeffrey Spaan, Kuan-Hsun Chen and Ana-Lucia Varbanescu [abstract] |
| 09:30 | Evaluating the Sustainability of Post-hoc XAI: A Comparative Study of LIME and SHAP
Daniel Amidirad, Bogdan Andrei, Ana Maria Oprescu and Sander Klous [abstract] |
| 09:45 | Application of dynamic modelling to the energy consumption of training neural networks
Maria Gonzalez Gutierrez, Ana Oprescu, Mª Ángeles Moraga, Félix García and Coral Calero [abstract] |
| 10:00 | Efficient and Introspective Synchronization for Network Digital Twins
Zhiheng Yang, Xiaoxuan Zhang, Adam Belloum, Paola Grosso and Chrysa Papagianni [abstract] |
| [10:15 - 10:45] Coffee Break |
|---|
| [10:45 - 11:30] Session 5: Distributed Systems, Services, and Security (Chair: Fernando Kuipers) |
|---|
| 10:45 | Bandwidth-aware Execution Time and Energy Predictor for Secure Inference Services
Tanjina Islam, Ana-Maria Oprescu, Zoltan Mann and Sander Klous [abstract] |
| 11:00 | Seeing Through NAT with Ripple: Retrofitting Service Dependency Discovery in Networked Distributed Systems
Diogo Landau, Gijs Blanken, Jorge Barbosa and Nishant Saurabh [abstract] |
| 11:15 | An Attribute-Driven Access Control Framework Based on Smart Contracts for Secure Collaborative Predictive Maintenance
Yago de R. dos Santos and Marcela Tuler de Oliveira [abstract] |
| [11:30 - 12:00] Coffee Break |
|---|
| [12:00 - 12:45] Session 6: Software Reliability and Program Analysis (Chair: Suzan Bayhan) |
|---|
| 12:00 | Correct-by-Construction MPI Codes (Early Ideas)
Badia Liokouras, Ben van Werkhoven and Rob V. van Nieuwpoort [abstract] |
| 12:15 | Uncovering Semantic Obstructions with LLVM Optimisation Remarks
Quinten Cabo, Kristian Rietveld and Rob van Nieuwpoort [abstract] |
| 12:30 | Finding Programming Faults Even When Large Parts of the Code have Disappeared
Quinten Cabo and Sven-Bodo Scholz [abstract] |
| [12:45 - 14:00] Lunch Break |
|---|
| [14:00 - 15:00] Keynote |
|---|
| 14:00 | Keynote: Misja Heuveling (IBM)
Succeeding with Agentic AI through Integration[abstract] |
| [15:00 - 15:15] Coffee Break |
|---|
| [15:15 - 16:15] Session 7: Edge AI and Efficient Inference Systems (Chair: Paola Grosso) |
|---|
| 15:15 | Elastoformer: Enabling Dynamic Adaptivity via Elastic Model Transformation
Sudaksh Kalra and Dolly Sapra [abstract] |
| 15:30 | Collaborative Inference in Battery-Powered Edge Networks
Mengyuan Li, George Iosifidis and Venkatesha Prasad [abstract] |
| 15:45 | Active Imitation Learning for Thermal- and Kernel-Aware LFM Inference on 3D S-NUCA Many-Cores
Yixian Shen, Andy Pimentel and Anuj Pathania [abstract] |
| 16:00 | Orthogonal Compression for Edge Vision Transformers: Combining Recursive Weight-Sharing with Token Merging
Junseo Kim, Uraz Odyurt and Amirreza Yousefzadeh [abstract] |
| [16:15 - 16:30] Coffee break |
|---|
| [16:30 - 17:30] Session 8: Heterogeneous Systems and Performance Analysis (Chair: Daniele Bonetta) |
|---|
| 16:30 | High-throughput JSONPath Query Execution on Neural Processing Units
Hexiang Geng, Tiziano De Matteis and Daniele Bonetta [abstract] |
| 16:45 | Exploring Streaming Time-Series-to-Graph Construction on Heterogeneous Platforms
Shaoshuai Du, Joze Rozanec, Ana Lucia Varbanescu and Andy D. Pimentel [abstract] |
| 17:00 | Roofplane: Analysing the Performance Bounds of Distributed Scientific Applications
Nihat Saritaş, Xavier Álvarez Farré and Ana-Lucia Varbanescu [abstract] |
| 17:15 | Safe Power Calculations in Multi-Core Processors Using Sub-Core Components
Derk Blom [abstract] |
| [17:30 - 18:00] Town Hall and Community Session (Chair: Steering Committee) |
|---|
| [18:00 - 20:30] Dinner (joint with NCCV) |
|---|
| [20:30] Pubquiz (joint with NCCV) |
|---|
| Day 3: Thursday June 18 |
| [9:00 - 10:00] Opening Session |
|---|
| 9:00 | Welcome
The CompSys 2026 organizers |
| 9:15 | Keynote: Zhiming Zhao (University of Amsterdam)
Quality-Critical Distributed Computing and Artificial Intelligence for Scientific Research [abstract] |
| [10:15 - 10:45] Coffee Break |
|---|
| [10:45 - 11:30] Session 9: Applied Machine Learning and Intelligent Sensing (Chair: Zhiming Zhao) |
|---|
| 10:45 | StemWin: Hybrid CNN-Transformer Architectures for Multiscale Crop Disease Detection
Horia Ionescu, Günes Özmen Bakan, Janik Euskirchen, Stan Ostaszewski, Dan Loznean, Vasile Mereuta, Marcin Pietrasik and Charalampos Kouzinopoulos [abstract] |
| 11:00 | Towards Device-Free Gaming with mmWave Radar
Yukuan Ding, Harvy Martinez, Girish Vaidya, Koen Langendoen and Marco Zuniga Zamalloa [abstract] |
| 11:15 | Autoencoders versus PCA for feature extraction in FDG PET scans in neurodegenerative diseases
Roland Veen, Sofie Lövdal, Kaitlin Vos, Ciro Setolino, Sanne Meles and Michael Biehl [abstract] |
| [11:30 - 11:45] Coffee Break |
|---|
| [11:45 - 12:00] Best Presentation Award |
|---|
| [12:00 - 12:45] Joint CompSys/NCCV Panel Discussion (Moderator: Giacomo D'amicantonio, TU/e) |
|---|
| 12:00 | Challenges and Opportunities in the dawn of Agentic AI era for Computer Systems and Computer Vision
|
| [12:45 - 13:00] Closing Session |
|---|
| 12:45 | Best ASCI Thesis Awards
Location: Tuinzaal |
2 - Keynotes
We have the pleasure of welcoming our keynote speakers at CompSys'26.
| Tuesday June 16, 10:00 |
| Title:
Emerging Industrial AI Systems: Large Industry Models, AI Infrastructure, and Quantum Computing
|
| By Andre Luckow (BMW Group IT & Ludwig-Maximilians-Universität München) |
| Abstract:
Computational demands in the automotive industry are rising rapidly due to the increasing use of AI across engineering, manufacturing, and business processes, as well as the growing scale of industrial and engineering simulation. In this talk, we present AI applications together with the compute and platform infrastructure required to support them. We focus on Large Industry Models (LIMs), which combine industrial data with domain-adapted foundation models to enable engineering reasoning and process-aware decision-making in industrial applications. While AI is transforming computing, continued scaling faces increasing computational and algorithmic limits. Thus, quantum computing may complement AI in selected domains. We therefore examine emerging quantum and hybrid algorithms for AI, optimization, and simulation, and discuss their integration into future heterogeneous compute architectures.
|
| Short bio:
Andre Luckow is Head of Innovation and Emerging Technologies at BMW Group IT and lecturer at Ludwig-Maximilians-Universität München. He leads BMW Group IT’s global Tech Office Network, with locations in the United States, China, India, and Germany. His research spans high-performance and distributed computing systems, including HPC and cloud middleware for heterogeneous compute infrastructures. More recently, his work has focused on AI/ML systems at scale and quantum-classical hybrid computing, with contributions to quantum middleware and hybrid algorithms for optimization and simulation.
|
| Wednesday June 17, 14:00 |
| Title:
Succeeding with Agentic AI through Integration
|
| By Misja Heuveling (IBM) |
| Abstract:
AI is a hot topic in boardrooms today. Today’s leaders see AI as the next Industrial Revolution, creating new opportunities. While exploring all the new opportunities, MIT reports that 95% of the AI projects never make it to production. During this session you will learn about one of the root causes for the failure of these projects, namely the IT complexity and difficulty to securely connect to the right data and systems. During this session we will explain what Agentic AI is and why AI agents only succeed when they are grounded in trusted, well-governed data and capabilities. Attendees will learn how integration platforms can securely expose enterprise data and business functions as reusable services, enabling AI agents to operate safely, reliably, and at scale. You will learn more about AI and MCP gateways, how these allow AI agents to communicate using the Model Context Protocol (MCP) through the integration platform to enterprise systems in such a way that AI projects can be taken into production in a successful manner.
|
| Short bio:
As Field CTO for IBMs integration portfolio, Misja Heuveling brings deep expertise in the integration domain, helping organizations navigate complex digital transformation challenges. Passionate about innovation, Misja is a trusted advisor to his customers across Benelux and beyond. Misja and his team partner with customers to architect scalable, future-proof integration strategies that bridge cloud and on-premises environments to accelerate business agility and productivity. His experience and strategic insight empower businesses to unlock the full potential of their data and applications.
|
| Thursday June 18, 09:15 |
| Title:
Quality-Critical Distributed Computing and Artificial Intelligence for Scientific Research
|
| By Zhiming Zhao (UvA) |
| Abstract:
Advanced data science and AI technologies are opening new pathways for addressing complex societal and scientific challenges. However, developing effective software systems to harness these opportunities involves significant software and computational hurdles. Integrating emerging technologies into the research lifecycle as a novel problem-solving paradigm requires assembling research software components with varying levels of maturity. Moreover, achieving the system-level performance demanded by quality-critical applications, such as Digital Twins, real-time simulations and predictive decision-making, necessitates end-to-end optimization spanning design, deployment, and runtime adaptation. AI technologies have attracted considerable attention for their role in developing algorithms tailored to such quality-critical scientific applications. In this talk, we explore AI-based methods for quality-critical programming, scheduling, and adaptation within a software framework that operates across multiple service layers, bridging both development and operations (DevOps). We also discuss the challenges that quality-critical computing research faces in supporting large-scale AI and Digital Twin applications across distributed infrastructures.
|
| Short bio:
Dr. Zhiming Zhao is an Associate Professor and leader of the Multiscale Networked Systems (MNS) research group at the Informatics Institute, University of Amsterdam (UvA). His research focuses on developing innovative programming and control models for quality-critical systems on programmable infrastructures — including Clouds, Edges, and Software-Defined Networks — leveraging optimization and artificial intelligence technologies. Supported by a range of EU and Dutch research projects, his team develops Digital Twinning solutions, Virtual Research Environments, and cloud automation tools to tackle data and computational challenges in both industrial innovation and scientific research. Dr. Zhao is a Senior Member of IEEE and serves as Managing Editor of the Journal of Cloud Computing.
|
3 - Panel Discussion
Challenges and Opportunities in the dawn of Agentic AI era for Computer Systems and Computer Vision
This joint panel aims to create a structured dialogue between the CompSys and NCCV communities on emerging research paradigms shaped by agentic AI. The session will explore how autonomous, goal-driven (agentic) AI systems are influencing research and educational methodologies, infrastructure, and evaluation practices across both communities, with a focus on identifying shared challenges and opportunities.
The key panelists will be:
- Christoph Lofi (TU Delft) - Christoph is an Associate Professor in the Web Information Systems group at Delft University of Technology and serves as the Director of Studies for the BSc Computer Science and Engineering programme. His long-term research vision is to develop semantic-based data and knowledge engineering methodologies that enable FAIR data management platforms to serve as a foundation for sustainable societal research. His work addresses research challenges related to knowledge extraction from unstructured data, dataset integration and enrichment, semantic query processing, and metadata management as a socio-technical system. Aligned with the university’s vision of “Impact for a Better Society,” Christoph’s research is inherently interdisciplinary, focusing on domains where innovative data engineering can address pressing societal needs. His primary application areas include agricultural and botanical sciences, public health, nutrition, and translational sciences.
- Andre Luckow (BMW Group IT & Ludwig-Maximilians-Universität München) - Andre Luckow is Head of Innovation and Emerging Technologies at BMW Group IT and lecturer at Ludwig-Maximilians-Universität München. He leads BMW Group IT’s global Tech Office Network, with locations in the United States, China, India, and Germany. His research spans high-performance and distributed computing systems, including HPC and cloud middleware for heterogeneous compute infrastructures. More recently, his work has focused on AI/ML systems at scale and quantum-classical hybrid computing, with contributions to quantum middleware and hybrid algorithms for optimization and simulation.
- Estefania Talavera (University of Twente) - Estefania is an Assistant Professor in Multimodal Learning for Human Behaviour Analysis within the Data Management and Biometrics group at the University of Twente. His research interests span computer vision, machine learning, and their intersection for understanding and analysing human behaviour. Prior to joining the University of Twente, he worked as a lecturer and researcher in the Information Systems group at the University of Groningen. He received his PhD in Computer Science in February 2020 from the University of Barcelona and the University of Groningen, under the supervision of Prof. Petia Radeva and Prof. Nicolai Petkov.
- Pascal Mettes (University of Amsterdam) - Pascal is a tenured Assistant Professor at the University of Amsterdam. Together with his research team, he aims to advance the field of hyperbolic deep learning. While contemporary deep learning is largely founded on Euclidean geometry, this framework has inherent limitations that cannot be overcome simply through larger datasets or more complex models. One of the most significant challenges concerns the representation and learning of hierarchical structures, which are fundamentally hyperbolic in nature due to their exponential growth patterns. His research focuses on developing the theoretical foundations and algorithmic methods required to perform deep learning in hyperbolic spaces, enabling more effective modelling of hierarchical data. His key research domains are highlighted below.
The moderator of the panel will be Giacomo D’Amicantonio (TU/e)
4 - Pictures
CompSys 2026 pictures
Coming Soon 🔜
5 - Town hall
This session is a status update on the CompSys community, and will offer opportunities to further connect to WGs, to contribute to the manifesto, and to brainstorm about ways the SC can help the community grow and thrive. We will keep the slides to a bare minimum and focus on discussions! Bring your thinking cap(s)!
The session is scheduled for Wednesday June 17, 17:30.
6 - Organization
Organization
Steering committee
| Name |
University/Organization |
| Paola Grosso | University of Amsterdam |
| Fernando A. Kuipers | TU Delft |
| Ana Lucia Varbanescu | University of Twente |
| Alexandru Iosup | Vrije Universiteit Amsterdam |
Technical Program Committee
| Name |
University/Organization |
| Anuj Pathania | University of Amsterdam |
| Jan Rellermeyer | Leibniz University Hannover |
| Kuan-Hsun Chen | University of Twente |
| Tiziano De Matteis | Vrije Universiteit Amsterdam |
| Nitinder Mohan | TU Delft |
| Mart Lubbers | Radboud University |
| Bernard van-Gastel | Radboud University |
| Remco Veltkamp | Utrecht University |
| Alessio Sclocco | Netherlands eScience Center |
| Savio Sciancalepore | Eindhoven University of Technology |
| Roopesh Kumar Polaganga | The University of Texas at Arlington |
| Leszek Ambroziak | Bialystok University of Technology, Faculty of Mechanical Engineering |
| Vineet Gokhale | Ghent University |
| Ben van Werkhoven | Leiden University |
| Kristian Rietveld | Leiden University |
| Kishor Joshi | TU Eindhoven |
| Nikolaos Alachiotis | University of Twente |
| Sven-Bodo Scholz | Radboud University |
| Lisa Maile | Technische Universität Braunschweig |
| Victoria Degeler | University of Amsterdam |
| Matthijs Jansen | Vrije Universiteit Amsterdam |
| Fernando Castor | University of Twente |
| Mitra Nasri | Eindhoven University of Technology |
| Tanya Shreedhar | TU Delft |
| Daniele Bonetta | VU Amsterdam |
| Erik van der Kouwe | Vrije Universiteit Amsterdam |
| Enkeleda Bardhi | Delft University of Technology |
| Mart Lubbers | Radboud University |
| Aske Plaat | Leiden University |
| Mengyuan Zhang | The Hong Kong Polytechnic University |
7 - Location
Location:
8 - Submission details
Submission Guidelines
All contributions will be reviewed by the Program Committee. Accepted
contributions will appear in the final program either as a short talk
or a full presentation, depending on the reviews. All presentations
will be made available in digital format, unless otherwise instructed
by the authors. For submission, the PDF format is mandatory.
To foster the broadest possible engagement and exchanging of ideas,
CompSys 2026 does not claim copyright, making it possible for authors
of accepted contributions to present work that has already been
published or is in the process of being published elsewhere.
Important dates
| Description |
Date |
| Paper submission deadline |
24 April 2026 |
| Author notification |
25 May 2026 |
| Registration deadline |
5 June 2026 |
| Conference dates |
16-18 June 2026 |
Submission Types
CompSys 2025 welcomes three types of contributions: research papers, work-in-progress papers/early ideas, and negative/failed research results.
Research/Long Papers
Research papers on your best research results from the past year(s). This includes papers already submitted to and/or accepted at (inter)national conferences or workshops.
Short Papers (work-in-progress/early ideas)
Since CompSys is a forum that encourages discussions about early and exciting ideas, we specifically welcome extended abstracts highlighting early ideas and work-in-progress papers. Such submissions are especially suitable for graduate and undergraduate students working towards finalizing their thesis or PhD students who have recently started or would like to share one of their preliminary results with the community. In particular, we encourage contributions in the form of short talks to share an early and not yet explored idea with the community to stimulate discussions and collect feedback. These talks might be particularly interesting for early-stage researchers. The paper should mention the research question being addressed, outline the novelty and/or originality of the idea, approach, or (initial) results, and contain a summary of preliminary results.
Negative/Failed Research Results
As in the previous years, we also solicit contributions sharing negative results, wrong methodologies, and/or invalidated hypothesis to share the lessons learned in the community and also once again remind to ourselves that a regular part of performing research is also about trying many ideas that may not lead to expected results.
Submission Guidelines
Research/Long papers
Long papers (not exceeding 12 pages in double-column or 15 pages in LNCS format) can be submitted using any of the commonly used templates (e.g., ACM, IEEE, LNCS).
Short Papers (work-in-progress/early ideas) and Negative/Failed Research Results
Submissions of early ideas, work-in-progress papers or negative/failed research results require a short paper of at most 2 pages (not including references) in IEEE double-column format or 4 pages (not including references) in LNCS single-column format.
Submission Portal
9 - Contact
International Conference on Computing Systems
Paula Diks (ASCI OFFICE)
Van Mourik Broekmanweg 6
2628 XE Delft
Netherlands
10 - Important Dates
| Description |
Date |
| Paper submission deadline |
10 May 2026 |
| Author notification |
25 May 2026 |
| Registration deadline |
5 June 2026 |
| Conference dates |
16-18 June 2026 |
| |
