Big Data 2025

Keynote 1
Maria Riveiro
Jönköping University

Title
Explaining AI: helpful, harmful, or just hype?

Visualization-Empowered Human-in-the-Loop Artificial Intelligence

Abstract
In the past years, experts in visualization, human-computer interaction and related fields have substantially contributed to the topic, for instance, by the development of visualization approaches to open the typically closed black box design of popular machine learning methods. However, the rapid developments in AI/ML potentially trigger a fundamental change in our understanding of the capabilities and applicability of the models as they are now also able to “interact” with the general population. What are the implications in terms of trust into the analytical results and potential biases that may occur? How should visualization research react and adapt to increase trust and call our attention to critical biases to avoid them?

Bio
Maria Riveiro received a Ph.D. degree in Computer Science from Örebro University, Sweden, in 2011. She is currently a Professor of Computer Science at the Department of Computer Science and Informatics at the School of Engineering, Jönköping University, with the Human-Centered Technology Group. Her main research interests are human-centered AI, explainable AI and visual analytics; she has worked designing, developing, and evaluating technologies that make our lives easier and better, with people in mind. Prof. Riveiro is the recipient of a starting and consolidator grant from the Swedish Research Council investigating how to tailor explanations from AI systems to users’ expectations and how to evaluate explainable AI systems.

Keynote 2
Fernando V. Paulovich
Eindhoven University of Technology

Title
Machine- or Human-centered Visual Analytics: Do we Still need Humans for Data Analysis?

Abstract
In response to the 2001 terrorist attacks in the United States, a research and development agenda was established to develop a new discipline that supports analytical reasoning for large databases, facilitated by interactive visual interfaces. So, Visual Analytics was born. As initially defined, Visual Analytics is the process where humans and machines cooperate to generate knowledge in a loop centered on the human to provide expert knowledge to the analytical process. Over the years, the acute need for various institutions, ranging from governmental to industry sectors, to shift their decisions to be more data-driven has transformed Visual Analytics into an essential tool, where the user serves as the "oracle" guiding the process. However, with the recent popularization of Generative Artificial Intelligence, particularly the popularization of Large Language Models, the central position of the expert human has started to be challenged, and one question is gaining traction: Can humans be replaced by intelligent (machine) agents in the analytical loop? In this talk, I will explore this new direction to raise awareness about possibilities and opportunities while pointing out intrinsic limitations.

Bio
Fernando V. Paulovich is an Associate Professor in Visual Analytics for Data Science at the Department of Mathematics and Computer Science, Eindhoven University of Technology (TU/e), the Netherlands. Before moving to the Netherlands, he was a Professor and Canada Research Chair at Dalhousie University, Canada (2017-2022), and an associate professor at the University of São Paulo, Brazil (2009-2017). He has been researching information visualization and visual analytics, focusing on integrating machine learning and visualization tools and techniques, taking advantage of the automation provided by machine learning methods and user knowledge gained through interactions with visual representations, enabling people to understand and utilize complex and massive data collections. In recent years, his primary focus has been on designing and developing visual analytics techniques for the general public to advance the concept of data democratization, promoting unconstrained access to data analysis and widening the analytic capability of lay users in transforming data into insights.

Keynote 3
Fredrik Heintz,
Linköping University

Title
Towards Trustworthy and Factual Large Language Models

Abstract
Europe has taken a clear stand that we want AI, but we do not want just any AI. We want AI that we can trust. This talk will present ongoing research from the EU project TrustLLM which has the goal of developing more factual and trustworthy large language models. To achieve the ambitious objectives of this project, TrustLLM will tackle the full range of challenges of LLM development, from ensuring sufficient quality and quantity of multilingual training data, to sustainable efficiency and effectiveness of model training, to enhancements and refinements for factual correctness, transparency, and trustworthiness, to a suite of holistic evaluation benchmarks validating the multi-dimensional objectives.

Bio
Fredrik Heintz is a Professor of Computer Science at Linköping University, where he leads the Division of Artificial Intelligence and Integrated Computer Systems (AIICS) and the Reasoning and Learning lab (ReaL). His research focus is artificial intelligence especially Trustworthy AI and the intersection between machine reasoning and machine learning. Director of the Wallenberg AI and Transformative Technologies Education Development Program (WASP-ED), Co-director of the Wallenberg AI, Autonomous Systems and Software Program (WASP), Coordinator of the TrustLLM project, and Vice President for AI Research Adra the AI, Data, and Robotics partnership. Member of the Swedish AI Commission. Fellow of the Royal Swedish Academy of Engineering Sciences (IVA).

Keynote 4
Masahiro Ryo, 
Brandenburg University of Technology

Title
Understanding biodiversity - Masahiro Ryo, Leibniz Centre for Agricultural Landscape Research

Abstract
Masahiro will present his research focusing on applying machine learning and explainable AI to biodiversity and sustainable agriculture. He develops innovative tools for scalable biodiversity monitoring and climate-resilient agriculture, often involving citizen science and collaboration with NPOs and companies In his talk, he will introduce ongoing projects that analyses stakeholders' opinions on sustainability using a large language model and facilitates biodiversity monitoring and education ("virtual ecologist").

Bio
Prof. Dr. Masahiro Ryo is Professor of Environmental Data Science at Brandenburg University of Technology and leads the AI group at the Leibniz Centre for Agricultural Landscape Research (ZALF) in Germany. Masahiro Ryo obtained a Ph.D. degree in Civil Engineering from Tokyo Institute of Technology, Japan in 2015. His research mission is to offer sustainable solutions for balancing nature and our society under global change. He studies the intersection of artificial intelligence, biodiversity conservation, and smart agriculture. Target scales range from microbial to global scales.  Prof. Ryo has published extensively and actively promotes impact-driven AI in environmental science.

Keynote 5
Laurynas Adomaitis, 
RISE - Research Institute of Sweden

Title
Ethics Readiness Levels: An Ethics-by-Design Approach for Data and AI

Abstract
It is a common practice in the European Union to frame ethical constraints on emerging technologies as “soft law”: codes of conduct, codes of practice, or guidelines. In domains of Data and Artificial Intelligence the “soft law” has given rise to “hard law” or regulation in GDPR and AI Act. However, the deontological approach based on values fails to grasp the complexity of developing technology, essentially trying to codify into high-level principles what are local, contextual, and embedded processes. As an alternative, I present an ethics-by-design methodology that uses a structured dialogue-based approach to 1) introduce ethics during the design of the system, as opposed to audits and standard conformity, which happens at then end, when the design is already finalized, 2) uses dialogue to elicit ethical reflection in a concrete use case setting, rather than working with abstract principles; and 3) produces a measurable track record of ethics maturity of a component over time. The ethics-by-design methodology is based on the idea of “ethics readiness”, structurally similar to the idea of “technological readiness levels.” It introduces ethical issues and concerns to the innovators in their own operating field and to guide their thinking in a systematic way. The dialogue on ethical issues involving different stakeholders should occur regularly and often recursively at different times and run in parallel with, and inform, the process of scientific research and technological design.

Bio
Laurynas Adomaitis is an AI Ethics and Governance Researcher at RISE Research Institutes of Sweden, specialising in bridging the gap between ethics and engineering practice. Previously, Laurynas was a postdoctoral researcher at CEA-Saclay, working in multiple EU Horizon projects. Laurynas has taught AI and Data Ethics at leading engineering faculties (SupOptique, CentraleSupélec) and business schools (emlyon) in Paris. He also has industry experience as an Innovation Manager at Nord Security, a cybersec unicorn from Vilnius. He defended his PhD in Philosophy cum laude at Scuola Normale Superiore in 2020.

Keynote 1
Maria Riveiro
Jönköping University

Title
Explaining AI: helpful, harmful, or just hype?

Visualization-Empowered Human-in-the-Loop Artificial Intelligence

Abstract
In the past years, experts in visualization, human-computer interaction and related fields have substantially contributed to the topic, for instance, by the development of visualization approaches to open the typically closed black box design of popular machine learning methods. However, the rapid developments in AI/ML potentially trigger a fundamental change in our understanding of the capabilities and applicability of the models as they are now also able to “interact” with the general population. What are the implications in terms of trust into the analytical results and potential biases that may occur? How should visualization research react and adapt to increase trust and call our attention to critical biases to avoid them?

Bio
Maria Riveiro received a Ph.D. degree in Computer Science from Örebro University, Sweden, in 2011. She is currently a Professor of Computer Science at the Department of Computer Science and Informatics at the School of Engineering, Jönköping University, with the Human-Centered Technology Group. Her main research interests are human-centered AI, explainable AI and visual analytics; she has worked designing, developing, and evaluating technologies that make our lives easier and better, with people in mind. Prof. Riveiro is the recipient of a starting and consolidator grant from the Swedish Research Council investigating how to tailor explanations from AI systems to users’ expectations and how to evaluate explainable AI systems.

Amilcar Soares
Linnaeus University

Title: Taxonomy-Driven Approaches for Exploring Movement Data

Abstract
The increasing availability of movement data—from ships and wildlife to storms and sports—offers unprecedented opportunities for understanding dynamic behaviors across domains. Yet this abundance also brings significant challenges: high dimensionality, lack of labels, and difficulty extracting interpretable patterns that can inform scientific hypotheses and practical applications. While valuable, traditional exploratory data analysis techniques often fall short in managing such complexity.
In this talk, I will present our efforts to address these challenges in two stages. First, we developed a theoretical framework that organizes movement variables into a structured taxonomy and leverages anomaly detection to reveal meaningful patterns in unlabeled, high-dimensional datasets. This provides a foundation for describing movement behaviors in an interpretable way. Building on this theoretical groundwork, we advanced toward practice by designing a visual analytics tool that operationalizes these ideas. The tool bridges statistical descriptors and interactive visualization, enabling analysts to explore movement data across multiple levels of abstraction.
Together, these contributions show how theory and tool development can be combined to advance our ability to describe and make sense of complex movement data, opening new avenues for exploratory analysis and knowledge discovery in diverse scientific and applied contexts.

Bio
Amilcar Soares is an Associate Professor in Computer Science at Linnaeus University, Sweden, and a member of the Centre for Data Intensive Sciences and Applications (DISA). Before joining Linnaeus University, he was an Assistant Professor at Memorial University of Newfoundland and a Research Associate at the Institute for Big Data Analytics, Dalhousie University, where he also held an Adjunct Professorship. He received his Ph.D. in Computer Science from the Federal University of Pernambuco, Brazil. His research focuses on spatiotemporal data analysis, trajectory mining, and visual analytics, with applications ranging from maritime traffic and ecology to sports and public health. He has contributed to several international research projects funded by agencies such as NSERC, DFO, Transport Canada, and DRDC Atlantic, and has published widely on methods combining machine learning and visualization for understanding complex movement data.

Keynote 2
Fernando V. Paulovich
Eindhoven University of Technology

Title
Machine- or Human-centered Visual Analytics: Do we Still need Humans for Data Analysis?

Abstract
In response to the 2001 terrorist attacks in the United States, a research and development agenda was established to develop a new discipline that supports analytical reasoning for large databases, facilitated by interactive visual interfaces. So, Visual Analytics was born. As initially defined, Visual Analytics is the process where humans and machines cooperate to generate knowledge in a loop centered on the human to provide expert knowledge to the analytical process. Over the years, the acute need for various institutions, ranging from governmental to industry sectors, to shift their decisions to be more data-driven has transformed Visual Analytics into an essential tool, where the user serves as the "oracle" guiding the process. However, with the recent popularization of Generative Artificial Intelligence, particularly the popularization of Large Language Models, the central position of the expert human has started to be challenged, and one question is gaining traction: Can humans be replaced by intelligent (machine) agents in the analytical loop? In this talk, I will explore this new direction to raise awareness about possibilities and opportunities while pointing out intrinsic limitations.

Bio
Fernando V. Paulovich is an Associate Professor in Visual Analytics for Data Science at the Department of Mathematics and Computer Science, Eindhoven University of Technology (TU/e), the Netherlands. Before moving to the Netherlands, he was a Professor and Canada Research Chair at Dalhousie University, Canada (2017-2022), and an associate professor at the University of São Paulo, Brazil (2009-2017). He has been researching information visualization and visual analytics, focusing on integrating machine learning and visualization tools and techniques, taking advantage of the automation provided by machine learning methods and user knowledge gained through interactions with visual representations, enabling people to understand and utilize complex and massive data collections. In recent years, his primary focus has been on designing and developing visual analytics techniques for the general public to advance the concept of data democratization, promoting unconstrained access to data analysis and widening the analytic capability of lay users in transforming data into insights.

Tora Hammar & Alisa Lincke 
Linnaeus University

Title
Using Large language models health care. Example of a research project to detect adverse drug events in clinical text 

Abstract
In the presentation we give an overview of possibilities and challenges of using large language models (LLMs) in medicine and health care. Then we present our ongoing project to fine tune and validate LLMs to detect adverse drug events (ADEs) in clinical texts. Effective methods to detect ADEs documented in an unstructured form will be valuable in increasing knowledge to prevent ADEs and improve patient safety.

Bio
Tora Hammar is an associate professor in health informatics, department of medicine and optometry in Kalmar. She works with research to improve medication safety and decision support, through better use of available health date. She is the research leader of the DISA group working with eHealth.

Alisa Lincke is Senior Lecturer at Linnaeus University, at the Department of Computer Science and Media Technology Faculty of Technology, in Växjö, Sweden. She works with data-driven applications, decision support systems, and machine learning, focused mainly on the healthcare sector.

Jonas Svensson
Linnaeus University

Title
Artificial ‘ulama - Analyzing AI-Generated Islamic Theology

Abstract
The presentation provides information on, and  preliminary findings from my research project Artificial ‘Ulama,examining how artificial intelligence systems produce Islamic theological content. The study focusses on how modern Large Language Models interpret and respond to prompts based on Islamic texts, concepts, and interpretational frameworks.

The investigation focuses on three domains: Qur'anic interpretation, religious counseling (fatwa), and contentious topics in Islamic faith and practice. Through systematic prompting with Islamic queries, I collect AI-generated theological responses and analyze the results through both distant and close reading techniques.

Our initial results reveal patterns of consistency and variation in AI theological approaches, offering insights into how algorithmic systems process Islamic knowledge. By comparing these outputs with established human interpretations, we identify convergences and divergences from traditional scholarly positions.

Bio
Jonas Svensson is professor in the Study of Religions at Linnaeus University. 

Keynote 3
Fredrik Heintz,
Linköping University

Title
Towards Trustworthy and Factual Large Language Models

Abstract
Europe has taken a clear stand that we want AI, but we do not want just any AI. We want AI that we can trust. This talk will present ongoing research from the EU project TrustLLM which has the goal of developing more factual and trustworthy large language models. To achieve the ambitious objectives of this project, TrustLLM will tackle the full range of challenges of LLM development, from ensuring sufficient quality and quantity of multilingual training data, to sustainable efficiency and effectiveness of model training, to enhancements and refinements for factual correctness, transparency, and trustworthiness, to a suite of holistic evaluation benchmarks validating the multi-dimensional objectives.

Bio
Fredrik Heintz is a Professor of Computer Science at Linköping University, where he leads the Division of Artificial Intelligence and Integrated Computer Systems (AIICS) and the Reasoning and Learning lab (ReaL). His research focus is artificial intelligence especially Trustworthy AI and the intersection between machine reasoning and machine learning. Director of the Wallenberg AI and Transformative Technologies Education Development Program (WASP-ED), Co-director of the Wallenberg AI, Autonomous Systems and Software Program (WASP), Coordinator of the TrustLLM project, and Vice President for AI Research Adra the AI, Data, and Robotics partnership. Member of the Swedish AI Commission. Fellow of the Royal Swedish Academy of Engineering Sciences (IVA).

Andreas Darnell
Södra Cell Technology Development

Title
Condition Based Monitoring at Scale 

Abstract
Condition-based monitoring of assets has long been employed in process industry for early detection of, possibly catastrophic failures, to protect assets and mitigate monetary loss. Traditionally online hard-wired vibration monitoring of a few key assets has been feasible. However, resent advancements in sensor technology, computational power and communication, have made sensors with higher sensitivity and bandwidth more affordable and accessible. With the development of long-range wireless communication large scale deployment of sensors has now become tractable. Södra Cell has made a strategic investment in large scale deployment of vibration monitoring sensors across all its production sites, with to goal of reaching an installed base of ~ 5.000 to 10.000 sensors. Typically, each asset would be assigned a unique set of monitoring characteristics which necessitates a judicious choice of computational methods, monitored frequencies, intensities and alarm thresholds. This highly specialized configuration is performed by a few, hard to come by, experts possessing deep knowledge within the specific field. Due to the noisy and intermittent environment, that makes up a process industry, continuous work on adjusting asset configurations is necessary, where factors such as intermittency and cross contamination between neighboring equipment plays a large role. As the number of deployed sensors continuous to increase, it continuously challenges the organization managing and monitoring the asset fleet. Thus, new and more automated ways to monitor and manage the ever-growing flood of information overwhelming the organization is called for. Together with Viking Analytics, Södra Cell has explored new ways of leveraging AI and ML to support and augment the human expert to support condition-based asset monitoring at scale.

Bio
Andreas Darnell is a Senior Development Engineering OT at Södra Cell Technology Development. His focus is advanced process control and supporting OT development and realization.

Felix Viberg
Linnaeus University and SKF Sverige AB

Title
Anomaly Dectection in Unlabeled Signals

Abstract
In this work, we propose a novel approach to anomaly detection utilizing normalizing flows, a type of deep generative model, and an adaptive learning scheme that implicitly labels samples based on their influence on the learning process. The method dynamically disregards samples whose loss significantly deviates from the mean during training, effectively treating them as noise or anomalies. We empirically demonstrate that such discarded samples tend to correspond to rare events and are hence anomalous. Our experimental results on synthetic datasets with varying levels of anomaly severity indicate that this method is robust and performs comparably to or better than a vanilla normalizing flow, even approaching the performance of a model trained solely on noise-free nominal data in many cases. We also evaluate the method on a real-world industrial dataset from bearing ring manufacturing, showing its capability to exclude unwanted noisy signals. Despite limitations in testing across diverse datasets like image data and the need for manual hyperparameter tuning, this approach offers a promising direction for robust anomaly detection in noisy, unlabeled industrial signals.

Bio
Felix Viberg, who obtained a master's degree in Engineering Physics from Chalmers University, is currently working towards a Ph.D. at LNU. With strong industry connections, Felix has engaged in applied research within manufacturing settings at SKF Sverige AB in Gothenburg. His research is concentrated on creating near real-time support systems for industrial equipment, such as PLCs, NC machines, robots, and CMMs.

Erdal Akin
Malmö University

Title
Deep Reinforcement Learning: From Foundations to Boundaries

Abstract
Deep Reinforcement Learning (DRL) has emerged as one of the most powerful paradigms in artificial intelligence, enabling agents to learn complex behaviors from raw sensory inputs through interaction with their environments. In this talk, we explore the foundational concepts of reinforcement learning and illustrate how deep learning extends its capabilities to high-dimensional problems. We review key DRL algorithms—such as DQN, PPO, and Actor-Critic methods—and examine their applications in robotics, autonomous systems, gaming, and cybersecurity. We also address the critical challenges in training DRL agents, including sample inefficiency and stability issues, and highlight emerging trends like multi-agent systems and offline RL. The presentation aims to provide a conceptual roadmap for the future of DRL in real-world decision-making systems.

Bio
Erdal Akin is an assistant professor at the Department of Computer Science and Media Technology, Malmö University, Sweden. He received his degree from the Department of Mathematics, Yildiz Technical University, Istanbul, in 2008 and his master’s and Ph.D. degrees from the Department of Computer Science, The University of Texas at San Antonio (UTSA), San Antonio, TX, USA, in 2014 and 2018, respectively. His research interests include deep reinforcement learning, software-defined networks, security, the IoT, and computer vision.

Björn Lindenberg
Linnaeus University

Title
Distributional Reinforcement Learning

Abstract
Distributional Reinforcement Learning (DRL) represents a recent and successful paradigm shift in reinforcement learning, especially for algorithms based on deep learning. Instead of estimating only expected returns, DRL agents learn the full distribution of possible outcomes, which admits a richer representation and greater flexibility for algorithm design. This approach has led to improved empirical performance in complex environments and enables new capabilities, such as risk-sensitive behavior.

Bio
Björn Lindenberg is a Senior Lecturer in Mathematics at Linnaeus University, specializing in machine learning. His doctoral work explored the intersection of dynamical systems and reinforcement learning. Currently, Björn's research interests are centered on developing novel and robust algorithms in reinforcement learning. He is currently involved in several research projects, including Algorithms for reinforcement learning and HPC for SME, which supports data-driven capabilities for regional enterprises, and a project on optimizing sustainable energy and food production in rural areas. He is a member of the research groups AI and machine learning for optimization and operations research and Computational Mathematics for Predictive Digital Twins (PreDiTwin).

Patrick Bergman,
Linnaeus University

Titel
The past present and future of wearables in health research

Abstract
Sensor-based assessments of physical activity have been available for over three decades. Initially, these technologies were primarily utilized by specialized researchers. However, in the past ten years, the consumer market for wearable devices—such as smartwatches and activity trackers—has expanded rapidly within the fitness industry. Recent innovations have improved the cost, size, and technical capability of wearables but their precision and transparency has not increased at a corresponding pace. Despite these limitations, the multifunctionality of wearables, akin to a "Swiss army knife," has sparked discussions about the potential of wearables as tools for self-monitoring and use in healthcare settings. Especially since they allow for the collection of long measurement periods of high-resolution information regarding both exposure and outcome simultaneously. In this presentation I will introduce the research area and describe some of the challenges that we must overcome so that in the future individuals may be able to improve and maintain their health through device-based services and make informed decisions based on their personal health data.

Bio
Patrick Bergman is an associate professor at the Department of Medicine and Optometry at Linnaeus University. He has over 20 years of experience in using sensor-based methods to assess physical activity behaviour with main area of interest in measurement errors associated with accelerometry, with an active interest in better predicting physical activity, especially the association between physical activity and health outcomes at the individual level.

Alisa Lincke,
Linnaeus University

Title
Validating Wearable Technologies for Blood Pressure Measurement: Challenges and Opportunities

Abstract
Hypertension affects nearly half of individuals over 65, yet current treatments often fail to achieve adequate blood pressure control.   Increased levels of physical activity (PA) have been shown to positively influence blood pressure and are recommended for hypertension treatment.  Evidence-based tools for modifying PA are lacking, and while wearables offer a promising solution for self-monitoring, only 11% have been validated — often with poor results. There is a need to validate and develop self-learning algorithms that can be implemented in wearables in order to suggest an appropriate level of PA based on blood pressure response to optimize blood pressure in patients with hypertension. The talk will share challenges and insights using wearable sensors to estimate systolic and diastolic blood pressure, focusing on data analysis from Bangle.js and EmotiBit devices.

Bio
Alisa Lincke is Senior Lecturer at Linnaeus University, at the Department of Computer Science and Media Technology Faculty of Technology, in Växjö, Sweden. She works with data-driven applications, decision support systems (DSSs), and machine learning, focused mainly on the healthcare sector.

Sara Caramaschi, 
Malmö University

Title
Ecological monitoring of physical activity from everyday data collected with digital health technology. Insights and significance for healthcare.

Abstract
With the advent of wearable devices and the widespread use of smartphones, people are increasingly collecting physical activity data from their everyday lives, collecting plenty of information, from steps, heart rate and heart rate variability, to measures of stress and energy levels. Movement and an active lifestyle can be considered as a prescription or intervention tool by doctors towards their patients. This happens in various conditions, such as cardiovascular or respiratory conditions, but also mental disorders and chronic diseases.

Despite the rich state-of-the-art of wearable devices and technology, there is a general lack of transparency with how metrics are computed, their accuracy and clinical relevance for healthcare.

This work investigates how physical activity measured during everyday life relates to standard physical capacity assessments such as the 6-Minute Walk Test and the Timed Up and Go Test. These tests are performed in today's traditional practices to investigate changes in patients' physical capacity, however, they have some shortcomings. They are usually performed with large time gaps from one hospital visit to the other, they struggle to be representative of natural symptom fluctuations and they have significant costs in terms of time and physical space for both patients and healthcare providers. Associating and inferring the test outcomes from everyday data has the potential to 1) provide more frequent assessment, 2) likely introduce a better representation of natural behaviour and activity levels, and 3) avoid high costs and travel times for clinicians and patients. In this context, transparent, reliable and actionable insights must be validated across populations, contexts and technology.

Bio
Sara (born in Mantova, Italy) is a biomedical engineer and researcher specializing in digital health and wearable technologies. She completed both her Bachelor's and Master's degrees in Biomedical Engineering at Politecnico di Milano, graduating in 2020 and 2022, respectively. During her studies, she participated in international academic experiences, including an Erasmus exchange at Oulu University in Finland, and a research internship at Philips Research in the Netherlands, where she worked on human activity recognition algorithms for her master’s thesis.

Currently, she is a Ph.D. candidate in Computer Science at Malmö University in Sweden. Her research focuses on the analysis of data from wearable devices and smartphones to assess health-related insights in real-life and unsupervised environments. She has contributed to publications in areas like human activity recognition, and physical capacity testing evaluation using wearable and mobile sensors.

Axel Djurberg,
Uppsala University

Title
Making Sense of Wearable Data: Physiology Meets Data Science

Abstract
As data from wearables—covering activity, sleep, and other daily metrics—continues to grow, new opportunities emerge to optimize health and performance at the individual level. At Svexa, we specialize in the development of data-driven “digital twins”—individualized models of physiology, recovery, and performance—used to simulate and translate complex data into actionable insights.This talk will explore how we leverage large, heterogeneous data sources and combine classical physiological models with machine learning pipelines to detect conditions such as overtraining and enhance training effectiveness.Many key physiological traits require laboratory-grade testing to assess accurately. At Svexa, we aim to infer such traits through data analysis—identifying patterns that allow us to estimate individual properties without the need for direct measurement. These patterns form the foundation of each person’s digital twin.

A major challenge in this space is data quality. Wearables can produce noisy or inconsistent measurements, and values can also vary significantly between sources. It will be discussed how these issues are tackled using a combination of data-driven individual modeling and domain-specific physiological expertise.This presentation aims to show how data science, when grounded in physiological understanding, can bridge the gap between data-driven insights and real-world variability in human data.

Bio
Axel holds a Master’s degree in Engineering Physics, with a specialization in data analysis and machine learning, from Uppsala University. At Svexa, he works across the full data pipeline—from investigating and parsing wearable data to performing analyses and developing software that delivers actionable insights. His work is grounded in close collaboration with Svexa’s physiology research team to ensure scientific rigor and domain relevance.
In addition to his technical role, Axel is an elite middle- and long-distance runner, with Swedish championship medals. His athletic background informs his interest in applying data science to sports and health contexts.

Jorge Zapico,
Linnaeus University

Title
Big Data and Sustainability: an overview

Abstract
In his talk Jorge will present an overview of the complex relationship between big data and sustainability based on his experiences with his previous work and research. This include direct negative impacts from the energy used and the production of the required hardware, and also indirect impacts as the technology is used, which can be positive if we are tackling a problem like saving energy, or negative if we are using technology to augment a problem. The presentation will provide some reflections about opportunity areas, risks and challenges and some points to think about in our role as technologists.

Bio
Jorge Zapico is a Senior Lecturer in Sustainable IT at LNU. He has an interdisciplinary background in computer science and sustainability, with his PhD at the Royal Institute of Technology in Stockholm exploring the use of data for working with sustainability problems. His projects have explored a variety of topics, from creating solar powered servers, visualizations for supermarkets to show how people buy organic food, always on displays for households with solar panels, to websites for municipalities to work with climate adaptation. In his research Jorge has explored both the possibilities of using data, but also identified risks, challenges and bias.

Keynote 4
Masahiro Ryo, 
Brandenburg University of Technology

Title
Understanding biodiversity - Masahiro Ryo, Leibniz Centre for Agricultural Landscape Research

Abstract
Masahiro will present his research focusing on applying machine learning and explainable AI to biodiversity and sustainable agriculture. He develops innovative tools for scalable biodiversity monitoring and climate-resilient agriculture, often involving citizen science and collaboration with NPOs and companies In his talk, he will introduce ongoing projects that analyses stakeholders' opinions on sustainability using a large language model and facilitates biodiversity monitoring and education ("virtual ecologist").

Bio
Prof. Dr. Masahiro Ryo is Professor of Environmental Data Science at Brandenburg University of Technology and leads the AI group at the Leibniz Centre for Agricultural Landscape Research (ZALF) in Germany. Masahiro Ryo obtained a Ph.D. degree in Civil Engineering from Tokyo Institute of Technology, Japan in 2015. His research mission is to offer sustainable solutions for balancing nature and our society under global change. He studies the intersection of artificial intelligence, biodiversity conservation, and smart agriculture. Target scales range from microbial to global scales.  Prof. Ryo has published extensively and actively promotes impact-driven AI in environmental science.

Graham Aid,
Ragn-Sells Group

Title
Accelerating Circular Scale-Up in a Volatile World

Abstract
Transforming sustainable lab innovations into full-scale industrial systems is rarely straightforward—especially in a world defined by volatile markets, complex material flows, and unpredictable disruptions. This talk will introduce a decision-first approach to scale-up, where digital twins and AI-powered scenario modeling help organizations accelerate innovation while managing uncertainty.

Drawing on experience from the environmental technology and waste valorization sector and recent projects like the Formas-funded PLENTY Center, Graham Aid shows how digital tools can tame the complexity of circular scaling—from variable feedstocks to dynamic downstream markets. This method prioritizes agility over infrastructure, using simulation and learning loops to make better choices early—before locking in costly commitments.

With roots in earlier architecture work from ISDATA.org, this presentation makes the case for data-first strategies in green innovation, highlighting how researchers and sustainability leaders alike can build resilience and fast-track promising ideas into profitable, scalable systems.

Bio
Graham Aid coordinates strategy and R&D at the Ragn-Sells Group, where he leads innovation in areas such as bioprocessing, hydrometallurgy, and detoxification. He holds a PhD in industrial symbiosis and innovation systems, with a focus on how emerging technologies can drive sustainability. His work bridges research and operations, using digital tools to scale circular solutions and transform complex industrial systems.

Johan Fransson,
Linnaeus University

Title
Single tree data for sustainable forestry

Abstract
Johan will present current advances in remote sensing and machine learning allowing a transition from stand based forestry management to single tree management . This higher definition of data can enable better monitoring of forest health, enable better continues cover forestry, and allow more heterogeneity and diversity in forest planning. Johan will present how new technologies like airborne lidar and hyperspectral cameras can be used to identify and measure single trees, and the data challenges and possibilities ahead.

Bio
Johan Fransson is a Professor in the Department of Forestry and Wood Technology at Linnaeus University (LNU), Sweden. Professor Fransson is actively involved in cutting-edge forestry research, serving as Research coordinator for the Forestry research group in Linnaeus University Centre for Data Intensive Sciences and Applications (DISA) since January 2022. His research focuses on the intersection of technology and sustainable forest management, with particular expertise in remote sensing applications for forestry. He leads several major research projects including OptiForValue, which focuses on optimising forest operations for sustainable forest management and high-value applications. OptiForValue is a Circular Bio-based Europe Joint Undertaking (CBE JU) research project that addresses challenges in European forestry by optimising forest-based value chains and enhancing forest management to foster more sustainable and resilient forest-based value chains. Professor Fransson has extensive international experience, having been a Senior Member of the Institute of Electrical and Electronics Engineers – Geoscience and Remote Sensing Society (IEEE-GRSS) since 2002 and a member of the Science Advisory Panel of the ALOS Kyoto & Carbon Initiative since 2004. His research contributes significantly to sustainable forest management by integrating advanced technologies with ecological and economic considerations.

Mauro Caporuscio & Samuele Giussani, 
Linnaeus University

Title
Architecting Carbon-aware Software-as-a-Service Applications

Abstract
Software-as-a-Service solutions are increasingly being adopted when developing software applications, as they are scalable, cost-effective, and facilitate rapid deployment while providing high availability and flexibility. However, the impact of Software-as-a-Service in terms of carbon emissions is not yet adequately addressed as a design concern, and most of the existing efforts revolve around measuring and containing the carbon impact after the deployment. This talk introduces a model-driven reasoning framework that integrates UML-based software architecture modeling with carbon-aware concerns. Architectural elements are supplemented with sustainability and performance properties of interest, and a model-driven transformation generates a simulation model to evaluate multiple architectural designs according to their Software Carbon Intensity and performance metrics. The results guide decision-making by assessing and comparing the trade-offs between performance and carbon intensity for the analyzed designs. In this way, the reasoning framework provides an automated, tool-supported approach to designing environmentally responsible Software-as-a-Service applications.

Bio
Samuele Giussani is a Ph.D. student at the Department of Computer Science and Media Technology at Linnaeus University, Växjö, Sweden. He obtained his bachelor’s and master’s degrees in Computer Science Engineering at Politecnico di Milano, Italy, in 2018 and 2021, respectively. His research focuses on evaluating and optimizing the carbon impact of software systems, with a particular interest in Software-as-a-Service (SaaS) applications. Other research interests include data visualization techniques, model-driven engineering and self-adaptive systems.

Helena Brandt, 
Fondia Legal Services

Title
Navigating the Ethical Minefield of AI: Legal Realities and Responsible Choices

Abstract
This talk will explore the evolving ethical challenges of AI—from bias and transparency to accountability and human oversight. Helena Brandt will offer insights on how businesses can make responsible choices today while preparing for tomorrow’s legal landscape.

Bio
Helena Brandt is AI & Privacy Lead, Senior Legal Counsel at Fondia Legal Services, where she specializes in AI, data protection, and tech regulation. With over a decade of experience advising scaleups and tech companies, Helena helps organizations navigate the legal and ethical complexities of emerging technologies. She combines a deep understanding of the European regulatory landscape – including the AI Act, GDPR, and related frameworks – with a pragmatic approach tailored to real-world innovation.

Keynote 5
Laurynas Adomaitis, 
RISE - Research Institutes of Sweden

Title
Ethics Readiness Levels: An Ethics-by-Design Approach for Data and AI

Abstract
It is a common practice in the European Union to frame ethical constraints on emerging technologies as “soft law”: codes of conduct, codes of practice, or guidelines. In domains of Data and Artificial Intelligence the “soft law” has given rise to “hard law” or regulation in GDPR and AI Act. However, the deontological approach based on values fails to grasp the complexity of developing technology, essentially trying to codify into high-level principles what are local, contextual, and embedded processes. As an alternative, I present an ethics-by-design methodology that uses a structured dialogue-based approach to 1) introduce ethics during the design of the system, as opposed to audits and standard conformity, which happens at then end, when the design is already finalized, 2) uses dialogue to elicit ethical reflection in a concrete use case setting, rather than working with abstract principles; and 3) produces a measurable track record of ethics maturity of a component over time. The ethics-by-design methodology is based on the idea of “ethics readiness”, structurally similar to the idea of “technological readiness levels.” It introduces ethical issues and concerns to the innovators in their own operating field and to guide their thinking in a systematic way. The dialogue on ethical issues involving different stakeholders should occur regularly and often recursively at different times and run in parallel with, and inform, the process of scientific research and technological design.

Bio
Laurynas Adomaitis is an AI Ethics and Governance Researcher at RISE Research Institutes of Sweden, specialising in bridging the gap between ethics and engineering practice. Previously, Laurynas was a postdoctoral researcher at CEA-Saclay, working in multiple EU Horizon projects. Laurynas has taught AI and Data Ethics at leading engineering faculties (SupOptique, CentraleSupélec) and business schools (emlyon) in Paris. He also has industry experience as an Innovation Manager at Nord Security, a cybersec unicorn from Vilnius. He defended his PhD in Philosophy cum laude at Scuola Normale Superiore in 2020.