Hotel Metropol, Belgrade
18th-22nd November 2024
DSC Europe 24 SCHEDULE
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.
Aristotle
Nuclear AI
Breakout Session
Description: Nuclear AI focuses on the transformative potential of artificial intelligence in the nuclear industry, from enhancing safety protocols to optimizing energy production. This session will explore how AI technologies are being integrated into nuclear power plants for predictive maintenance, real-time monitoring, and risk assessment. Attendees will also learn about the role of AI in advancing nuclear research, ensuring safer handling of materials, and improving decision-making processes in high-stakes environments. Join us to discover how AI is driving innovation and efficiency in the nuclear sector.
10:00 – 10:15 AI in Nuclear Knowledge Management
Riccardo Cocci
Product Manager @ Ask for the moon
At Ask for the Moon, we recognize that expertise is the key to driving business performance. However, much of this knowledge—especially tacit know-how gathered from experience—is hard to access, often trapped in employees’ minds. Workers lose up to three hours daily searching for answers. Our AI-powered platform bridges this gap by connecting those who possess knowledge with those who need it. When a question is posed, our AI notifies the right experts within the organization, enabling them to share their expertise quickly, accelerating projects, and ensuring that knowledge circulates and benefits everyone.
10:15 – 10:30 Prediction models used for maintenance & sales in nuclear power plant
Peter Kertys
Head of Data Science @ Slovenské elektrárne
Short introduction of deployed simple AI solutions in nuclear power plant without safety impact. Presented usecases inlcude approach to predictive maintenace via data collection, model creation and self learning setup. Additionaly power prediction models based on weather forecast used for sales would be presented.
10:30 – 10:45 TBA
Jovana Knezevic
Research Associate @ Faculty of Sciences, University of Novi Sad
TBA
11:00 – 12:00 Panel duscussion
Moderator: Jovana Knezevic
Research Associate @ Faculty of Sciences, University of Novi Sad
10:00 – 10:15 AI in Nuclear Knowledge Management
Riccardo Cocci
Product Manager @ Ask for the moon
At Ask for the Moon, we recognize that expertise is the key to driving business performance. However, much of this knowledge—especially tacit know-how gathered from experience—is hard to access, often trapped in employees’ minds. Workers lose up to three hours daily searching for answers. Our AI-powered platform bridges this gap by connecting those who possess knowledge with those who need it. When a question is posed, our AI notifies the right experts within the organization, enabling them to share their expertise quickly, accelerating projects, and ensuring that knowledge circulates and benefits everyone.
10:15 – 10:30 Prediction models used for maintenance & sales in nuclear power plant
Peter Kertys
Head of Data Science @ Slovenské elektrárne
Short introduction of deployed simple AI solutions in nuclear power plant without safety impact. Presented usecases inlcude approach to predictive maintenace via data collection, model creation and self learning setup. Additionaly power prediction models based on weather forecast used for sales would be presented.
10:30 – 10:45 TBA
Jovana Knezevic
Research Associate @ Faculty of Sciences, University of Novi Sad
TBA
11:00 – 12:00 Panel duscussion
Moderator: Jovana Knezevic
Research Associate @ Faculty of Sciences, University of Novi Sad
10:00
–
12:00
Nuclear AI
Riccardo Cocci & Peter Kertys & Jovana Knezevic
Product Manager @ Ask for the moon & Head of Data Science @ Slovenské elektrárne & Research Associate @ Faculty of Sciences, University of Novi Sad
Ethical Considerations in Data Science and AI
Breakout Session
Description: Ethical Considerations in Data Science and AI addresses the critical challenges surrounding the responsible development and deployment of AI technologies. This session will explore topics such as the preservation of linguistic diversity in the age of large language models (LLMs) and the ethical implications of biased training datasets. We will also examine global legal frameworks governing AI, with a focus on balancing innovation with ethical oversight. Attendees will gain insights into how developers, policymakers, and researchers can collaborate to ensure AI systems are both inclusive and aligned with societal values.
12:00 – 12:15 A Duty to Digitally Preserve Linguistic Diversity in the Age of LLMs
Vanja Subotic
Postdoctoral Researcher @ University of Belgrade – Faculty of Philosophy
The quality of training datasets for LLMs varies significantly across different natural languages. I focus on the challenge related to the preservation of cultural heritage: LLMs struggle to fully meet the linguistic needs of ethnic communities speaking low-resource languages, as chatbots based on these models are often less efficient and inclusive. This issue has broader implications: unless collective action is taken to align LLM performance with principles of linguistic diversity and justice, significant cultural nuances risk being lost through cultural appropriation. I propose a theoretical framework assigning responsibilities to developers, decision-makers, and academics on how to digitally preserve linguistic diversity. I will also outline practical steps for achieving this, drawing on the results of workshops I conducted to evaluate ChatGPT’s performance on Serbian literary texts, as Serbian is also considered a low-resource language.
12:15 – 12:30 Legal and ethical challenges in contemporary AI development and how to deal with them
Dusan Pavlovic
Legal and Ethics Expert – Data Protection and AI
The talk topic presents the latest regulatory development of AI and the presentation of the most significant regulatory requirements. The focus will be on Europe (EU), without neglecting current developments in the US, Canada, China, and the rest of the world. By highlighting recently enacted regulations, the talk attempts to outline similarities and differences in the regulation of AI on a global level. Ethical aspects of regulation could show what the greatest social concerns are and how to mitigate them. This will result in recommendations for building appropriate AI governance mechanisms that can ensure the safe development and deployment of AI but also increase its effectiveness.
12:30 – 12:45 The Power and Responsibility of Machine Learning: Ethical and Responsible AI in Action
Sray Agarwal
Director Responsible AI @ Fractal
Over the past decade, machine learning has become ubiquitous, transforming our lives in ways that were once unimaginable. Machines now make decisions that used to require significant human effort, but at much faster speeds and reduced costs—with minimal human oversight. As a result, AI systems are not only shaping our lives more than ever before but are also facing increased scrutiny from regulators and user rights advocates. The saying “with great power comes great responsibility” has been echoed throughout history, from the French Revolution to superhero comics. Today, this adage has never been more relevant. The immense power of machine learning is now accessible to anyone developing software, influencing everything from financial access that can alter someone’s life trajectory, to medical diagnoses that could extend or reduce life expectancy. Even the social media content we consume, driven by AI algorithms, can engage us or polarize us by reinforcing existing beliefs. In this context, it is crucial to approach AI through the lens of Responsible and Ethical AI (RAI). This involves focusing on key principles such as: Fairness: Ensuring discrimination-free algorithms. Explainability: Making AI decisions understandable from data, model, business, risk, and counterfactual perspectives. Privacy: Incorporating privacy by design in both systems and algorithms. Accountability: Holding AI systems and developers accountable for their outcomes. Sustainability: Developing AI systems that are environmentally and socially sustainable. This talk will highlight real-world examples from various domains, providing actionable strategies for implementing Responsible AI. It will also explore how these principles extend to the rapidly evolving world of Generative AI (like ChatGPT, Bard, etc.), ensuring that the next generation of AI systems is both powerful and ethical.
12:45 – 13:00 Deepfakes and the Law: The Battle for Trust in the Age of AI
Tijana Zunic Maric
Attorney at Law – Partner @ Zunic Law Firm
AI-generated content is reshaping how personal and corporate reputations are managed online, posing challenges in distinguishing real from fake. UCL has ranked deepfakes as the most serious AI-related crime threat among all forms of AI-generated content. Its impact spans issues from image-based sexual abuse to political manipulation. This session will explore the rise of deepfakes, its applications, legal implications in the EU, challenges for digital businesses, and strategies for mitigating risks and managing impacts considering recent legislation, including the EU Digital Services Act and the EU AI Act.
13:00 – 14:00 Panel duscussion
Moderator: Jelena Djukanovic
Attorney At Law @ Zunic Law Firm
12:00 – 12:15 A Duty to Digitally Preserve Linguistic Diversity in the Age of LLMs
Vanja Subotic
Postdoctoral Researcher @ University of Belgrade – Faculty of Philosophy
The quality of training datasets for LLMs varies significantly across different natural languages. I focus on the challenge related to the preservation of cultural heritage: LLMs struggle to fully meet the linguistic needs of ethnic communities speaking low-resource languages, as chatbots based on these models are often less efficient and inclusive. This issue has broader implications: unless collective action is taken to align LLM performance with principles of linguistic diversity and justice, significant cultural nuances risk being lost through cultural appropriation. I propose a theoretical framework assigning responsibilities to developers, decision-makers, and academics on how to digitally preserve linguistic diversity. I will also outline practical steps for achieving this, drawing on the results of workshops I conducted to evaluate ChatGPT’s performance on Serbian literary texts, as Serbian is also considered a low-resource language.
12:15 – 12:30 Legal and ethical challenges in contemporary AI development and how to deal with them
Dusan Pavlovic
Legal and Ethics Expert – Data Protection and AI
The talk topic presents the latest regulatory development of AI and the presentation of the most significant regulatory requirements. The focus will be on Europe (EU), without neglecting current developments in the US, Canada, China, and the rest of the world. By highlighting recently enacted regulations, the talk attempts to outline similarities and differences in the regulation of AI on a global level. Ethical aspects of regulation could show what the greatest social concerns are and how to mitigate them. This will result in recommendations for building appropriate AI governance mechanisms that can ensure the safe development and deployment of AI but also increase its effectiveness.
12:30 – 12:45 The Power and Responsibility of Machine Learning: Ethical and Responsible AI in Action
Sray Agarwal
Director Responsible AI @ Fractal
Over the past decade, machine learning has become ubiquitous, transforming our lives in ways that were once unimaginable. Machines now make decisions that used to require significant human effort, but at much faster speeds and reduced costs—with minimal human oversight. As a result, AI systems are not only shaping our lives more than ever before but are also facing increased scrutiny from regulators and user rights advocates. The saying “with great power comes great responsibility” has been echoed throughout history, from the French Revolution to superhero comics. Today, this adage has never been more relevant. The immense power of machine learning is now accessible to anyone developing software, influencing everything from financial access that can alter someone’s life trajectory, to medical diagnoses that could extend or reduce life expectancy. Even the social media content we consume, driven by AI algorithms, can engage us or polarize us by reinforcing existing beliefs. In this context, it is crucial to approach AI through the lens of Responsible and Ethical AI (RAI). This involves focusing on key principles such as: Fairness: Ensuring discrimination-free algorithms. Explainability: Making AI decisions understandable from data, model, business, risk, and counterfactual perspectives. Privacy: Incorporating privacy by design in both systems and algorithms. Accountability: Holding AI systems and developers accountable for their outcomes. Sustainability: Developing AI systems that are environmentally and socially sustainable. This talk will highlight real-world examples from various domains, providing actionable strategies for implementing Responsible AI. It will also explore how these principles extend to the rapidly evolving world of Generative AI (like ChatGPT, Bard, etc.), ensuring that the next generation of AI systems is both powerful and ethical.
12:45 – 13:00 Deepfakes and the Law: The Battle for Trust in the Age of AI
Tijana Zunic Maric
Attorney at Law – Partner @ Zunic Law Firm
AI-generated content is reshaping how personal and corporate reputations are managed online, posing challenges in distinguishing real from fake. UCL has ranked deepfakes as the most serious AI-related crime threat among all forms of AI-generated content. Its impact spans issues from image-based sexual abuse to political manipulation. This session will explore the rise of deepfakes, its applications, legal implications in the EU, challenges for digital businesses, and strategies for mitigating risks and managing impacts considering recent legislation, including the EU Digital Services Act and the EU AI Act.
13:00 – 14:00 Panel duscussion
Moderator: Jelena Djukanovic
Attorney At Law @ Zunic Law Firm
12:00
–
14:00
Ethical Considerations in Data Science and AI
Vanja Subotic & Dusan Pavlovic & Sray Agarwal & Tijana Zunic Maric & Jelena Djukanovic
Postdoctoral Researcher @ University of Belgrade – Faculty of Philosophy & Legal and Ethics Expert – Data Protection and AI & Director Responsible AI @ Fractal & Attorney at Law – Partner @ Zunic Law Firm & Attorney At Law @ Zunic Law Firm
AI in EdTech
Breakout Session
Description: AI is not just changing the way we learn—it’s reshaping the entire educational landscape and beyond. This session will take you on a journey through the powerful role AI plays in personalizing education, enhancing student engagement, and streamlining assessments. As AI evolves, its influence reaches far beyond traditional classrooms, transforming non-technical professions and offering new pathways for productivity across industries.
We will explore how AI is preparing the next generation of experts, not just in technical fields but also in interdisciplinary areas essential for the digital economy. You’ll hear about innovative educational models that blend AI with management and economics, creating professionals ready to navigate the complexities of a digital future.
Finally, we’ll look at AI’s role in academic research, particularly through generative AI tools that are redefining how scientific writing is done, speeding up the process while maintaining quality. Throughout this session, we’ll tackle the ethical questions and inclusivity challenges that come with integrating AI into education, ensuring we harness its power responsibly for a better, more equitable future.
15:00 – 15:15 AI in EdTech: Personalizing Learning, Assessing at Scale, and Boosting Engagement
Luka Anicin
CEO @ Datablooz
Explore how AI is transforming education by enabling personalized learning experiences, automating assessments, and fostering student engagement. This session will delve into practical examples of AI-driven platforms, tools, and strategies that help educators create more tailored and efficient learning environments. Attendees will gain insights into how AI can enhance curriculum delivery, streamline grading, and provide real-time feedback, while addressing ethical concerns and promoting inclusive education for diverse learners. Discover actionable ways to implement AI in your educational setting for better outcomes.
15:15 – 15:30 Lessons from the Trenches: AI as a Teaching Assistant and a Curriculum Game-Changer
Nebojsa Vasiljevic
Director of AI @ Fondacija Petlja
Our learning platform, Petlja.org, offers free and open online courses covering most IT-related subjects for K-12 education in Serbia. Over the past two years, we have explored several directions for developing AI on our platform and within our content. Ultimately, we focused on two key initiatives: developing an AI assistant for teachers and launching a new series of online courses, Instructor, designed to integrate AI into the curriculum. The AI assistant for teachers is entering its final pilot phase, and the first course in the Instructor series is set to be released soon. This talk will highlight our most intriguing findings and the open questions we continue to explore.
15:30 – 15:45 New Education Perspectives – How to Profile Experts for a Digital Economy Era
Ivan Lukovic
Full Professor @ University of Belgrade – Faculty of Organizational Sciences
Nowadays, we face a lack of strongly educated and interdisciplinary oriented experts showing an appropriate level of knowledge both in Computer Science, Software Engineering, and Artificial Intelligence, as well as in the disciplines of Management and Economics, for a specific problem domain. In this talk, we address issues on how to come to more flexible and interdisciplinary oriented study models capable of producing various forms of digital managers, as a new profile of experts, ready to cope with digital economy and digital transformation in a modern society. Massive deployment of such experts is a way to significantly raise the level of organization maturity regarding capabilities for: information management, quality management, and big data analytics.
15:45 – 16:00 Hello Claude. I am a scientist, how can you help me?
Dijana Oreski
Associate Professor @ University of Zagreb, Faculty of Organization and Informatics
This talk explores the potential of generative artificial intelligence (genAI) in higher education, with a particular focus on its applications in scientific writing. As genAI technologies become increasingly sophisticated, they offer various opportunities to enhance research productivity. The presentation will cover: (i) an overview of current AI tools for scientific writing, including literature reviews, drafting manuscripts, and editing, (ii) the importance of prompt engineering in effectively using genAI tools for scientific tasks.
16:00 – 17:00 Panel duscussion
Moderator: Ivan Lukovic
Full Professor @ University of Belgrade – Faculty of Organizational Sciences
We will explore how AI is preparing the next generation of experts, not just in technical fields but also in interdisciplinary areas essential for the digital economy. You’ll hear about innovative educational models that blend AI with management and economics, creating professionals ready to navigate the complexities of a digital future.
Finally, we’ll look at AI’s role in academic research, particularly through generative AI tools that are redefining how scientific writing is done, speeding up the process while maintaining quality. Throughout this session, we’ll tackle the ethical questions and inclusivity challenges that come with integrating AI into education, ensuring we harness its power responsibly for a better, more equitable future.
15:00 – 15:15 AI in EdTech: Personalizing Learning, Assessing at Scale, and Boosting Engagement
Luka Anicin
CEO @ Datablooz
Explore how AI is transforming education by enabling personalized learning experiences, automating assessments, and fostering student engagement. This session will delve into practical examples of AI-driven platforms, tools, and strategies that help educators create more tailored and efficient learning environments. Attendees will gain insights into how AI can enhance curriculum delivery, streamline grading, and provide real-time feedback, while addressing ethical concerns and promoting inclusive education for diverse learners. Discover actionable ways to implement AI in your educational setting for better outcomes.
15:15 – 15:30 Lessons from the Trenches: AI as a Teaching Assistant and a Curriculum Game-Changer
Nebojsa Vasiljevic
Director of AI @ Fondacija Petlja
Our learning platform, Petlja.org, offers free and open online courses covering most IT-related subjects for K-12 education in Serbia. Over the past two years, we have explored several directions for developing AI on our platform and within our content. Ultimately, we focused on two key initiatives: developing an AI assistant for teachers and launching a new series of online courses, Instructor, designed to integrate AI into the curriculum. The AI assistant for teachers is entering its final pilot phase, and the first course in the Instructor series is set to be released soon. This talk will highlight our most intriguing findings and the open questions we continue to explore.
15:30 – 15:45 New Education Perspectives – How to Profile Experts for a Digital Economy Era
Ivan Lukovic
Full Professor @ University of Belgrade – Faculty of Organizational Sciences
Nowadays, we face a lack of strongly educated and interdisciplinary oriented experts showing an appropriate level of knowledge both in Computer Science, Software Engineering, and Artificial Intelligence, as well as in the disciplines of Management and Economics, for a specific problem domain. In this talk, we address issues on how to come to more flexible and interdisciplinary oriented study models capable of producing various forms of digital managers, as a new profile of experts, ready to cope with digital economy and digital transformation in a modern society. Massive deployment of such experts is a way to significantly raise the level of organization maturity regarding capabilities for: information management, quality management, and big data analytics.
15:45 – 16:00 Hello Claude. I am a scientist, how can you help me?
Dijana Oreski
Associate Professor @ University of Zagreb, Faculty of Organization and Informatics
This talk explores the potential of generative artificial intelligence (genAI) in higher education, with a particular focus on its applications in scientific writing. As genAI technologies become increasingly sophisticated, they offer various opportunities to enhance research productivity. The presentation will cover: (i) an overview of current AI tools for scientific writing, including literature reviews, drafting manuscripts, and editing, (ii) the importance of prompt engineering in effectively using genAI tools for scientific tasks.
16:00 – 17:00 Panel duscussion
Moderator: Ivan Lukovic
Full Professor @ University of Belgrade – Faculty of Organizational Sciences
15:00
–
17:00
AI in EdTech
Luka Anicin & Nebojsa Vasiljevic & Ivan Lukovic & Dijana Oreski
CEO @ Datablooz & Director of AI @ Fondacija Petlja & Full Professor @ University of Belgrade – Faculty of Organizational Sciences & Associate Professor @ University of Zagreb, Faculty of Organization and Informatics
Plato
HPC & AI
Breakout Session
Description: HPC & AI explores the intersection of high-performance computing (HPC) and artificial intelligence, highlighting how these technologies are working together to accelerate advancements across industries. This session will cover how HPC systems are enabling more complex AI models and simulations, optimizing large-scale data processing, and driving breakthroughs in fields like scientific research and engineering. Attendees will gain insights into how HPC infrastructure is transforming AI capabilities, making it possible to solve problems that were previously computationally infeasible.
10:00 – 10:15 High Performance Computing Education: Empowering Engineers to Enable the AI Revolution
Dusan Gajic
Associate Professor @ Faculty of Technical Sciences, University of Novi Sad
High performance computing (HPC) has long been the driver behind scientific innovation, and now, with the AI revolution, it is become a key driver in how we store, process, and analyze data in science, business, and personal life. In this talk we’ll offer an insight into how we currently teach computer science master students to create efficient HPC solutions for real-world problems. Then we’ll discuss a new initiative which aims to address the topic of HPC education on the European level.
10:15 – 10:30 User friendly HPC: AI on HPC cluster for researchers and startups
Dusan Simic
Teaching associate @ Chair of Computer Science, Faculty of Sciences
An overview of a general purpose HPC cluster with an aim to offer user friendly approach to training AI models and other general computing tasks. We will review the need for a combination of proper tools, support and hardware/software compatibility in any emerging technology for greater user adoption.
10:30 – 10:45 TBA
Milutin Brankovic
Researcher @ NextSilicon
TBA
10:45 – 11:00 Foundation Models of Compiler Optimization
Dejan Grubisic
Research Scientist @ Meta
Link for the ABSTRACT
11:00 – 12:00 Panel duscussion
Moderator: Jelena Erakovic
Director of Marketing @ NextSilicon
10:00 – 10:15 High Performance Computing Education: Empowering Engineers to Enable the AI Revolution
Dusan Gajic
Associate Professor @ Faculty of Technical Sciences, University of Novi Sad
High performance computing (HPC) has long been the driver behind scientific innovation, and now, with the AI revolution, it is become a key driver in how we store, process, and analyze data in science, business, and personal life. In this talk we’ll offer an insight into how we currently teach computer science master students to create efficient HPC solutions for real-world problems. Then we’ll discuss a new initiative which aims to address the topic of HPC education on the European level.
10:15 – 10:30 User friendly HPC: AI on HPC cluster for researchers and startups
Dusan Simic
Teaching associate @ Chair of Computer Science, Faculty of Sciences
An overview of a general purpose HPC cluster with an aim to offer user friendly approach to training AI models and other general computing tasks. We will review the need for a combination of proper tools, support and hardware/software compatibility in any emerging technology for greater user adoption.
10:30 – 10:45 TBA
Milutin Brankovic
Researcher @ NextSilicon
TBA
10:45 – 11:00 Foundation Models of Compiler Optimization
Dejan Grubisic
Research Scientist @ Meta
Link for the ABSTRACT
11:00 – 12:00 Panel duscussion
Moderator: Jelena Erakovic
Director of Marketing @ NextSilicon
10:00
–
12:00
HPC & AI
Dusan Gajic & Dusan Simic & Milutin Brankovic & Dejan Grubisic & Jelena Erakovic
Associate Professor @ Faculty of Technical Sciences, University of Novi Sad & Teaching associate @ Chair of Computer Science, Faculty of Sciences & Researcher @ NextSilicon & Research Scientist @ Meta
Data-Driven Energy Management
Breakout Session
Description: Data-driven energy management is transforming the way we understand and optimize energy systems. In this session, we begin by examining how deep learning and IoT devices enable non-intrusive load monitoring, providing insights into household energy consumption and disaggregation. Next, we explore the application of explainable AI in district heating, focusing on how transparency and trust can be built into predictive models that guide energy system controls. We then shift focus to the future of mobility, showcasing how data science is revolutionizing battery development for electric vehicles. Finally, we conclude with an innovative use case of drone technology for inspecting power lines and energy networks, illustrating how these advancements streamline maintenance processes and improve energy infrastructure efficiency. Join us to explore how data science is shaping the future of energy management across various domains.
12:00 – 12:15 Non-Intrusive Load Monitoring Using IoT Smart Devices and Deep Learning
Francesco Conti
Data Scientist @ Agile Lab
In data-driven energy management, Non-Intrusive Load Monitoring (NILM) is a key method for identifying the energy use of individual household appliances. By enabling detailed tracking of appliance-level energy consumption, NILM supports improved energy efficiency and cost savings. This talk: 1. Introduces NILM, its benefits, and current state-of-the-art machine learning techniques; 2. Discusses smart IoT devices for NILM, with a focus on the Chain2 protocol implemented in Italian energy systems. This event-based protocol produces low-volume data, simplifying NILM implementation; 3. Shows how deep learning models can be trained using data from smart IoT devices.
12:15 – 12:30 Explainable AI-driven heat demand forecasting for transparent and trustworthy district heating systems control
Milan Zdravkovic
Data scientist and associated professor @ Faculty of Mechanical Engineering, University of Nis
Once you step aside from the glorified landscape of generative AI with its perceived hyper-realism and hypnotizing illusion of creativity, you find yourself in the world of automation, THE actual billion-dollar holy grail of the artificial intelligence. In this talk, I will show how to use AI to solve industrial control automation problem by relying on the forecasts of heat demand in residential areas. How Explainable AI can make this solution trustful? How to interpret different kinds of feature importances? What are the anchors, prototypes and counterfactual explanations and how we can use those to deconstruct empirical, data-driven into formal knowledge which can be understood and validated by district heating experts?
12:30 – 12:45 ET: Observing the Future
Aleksandar Lukic
Machine Learning Engineer @ Planet Soft
ET doesn’t stand for Extra-Terrestrial, but rather for Emerging Technology, Efficiency Toolkit, Enhanced Telemetry, and Enterprise Transformation. This system represents a new era in AI-driven solutions for energy infrastructure, specifically designed to transform how high-voltage power lines are inspected and maintained. ET leverages drone-collected LIDAR, RGB, and thermal data, processed through advanced machine learning and data analytics, to enable utility companies to proactively identify and address potential issues. This continuous, intelligent analysis allows for the precise detection of malfunctions and anomalies, reinforcing the stability and security of Europe’s energy grid. ET is more than a system—it’s a vision of a resilient and forward-thinking energy future.
13:00 – 14:00 Panel duscussion
Moderator: Lisa Kratochwill
Startup Engagement Manager @ VERBUND AG
12:00 – 12:15 Non-Intrusive Load Monitoring Using IoT Smart Devices and Deep Learning
Francesco Conti
Data Scientist @ Agile Lab
In data-driven energy management, Non-Intrusive Load Monitoring (NILM) is a key method for identifying the energy use of individual household appliances. By enabling detailed tracking of appliance-level energy consumption, NILM supports improved energy efficiency and cost savings. This talk: 1. Introduces NILM, its benefits, and current state-of-the-art machine learning techniques; 2. Discusses smart IoT devices for NILM, with a focus on the Chain2 protocol implemented in Italian energy systems. This event-based protocol produces low-volume data, simplifying NILM implementation; 3. Shows how deep learning models can be trained using data from smart IoT devices.
12:15 – 12:30 Explainable AI-driven heat demand forecasting for transparent and trustworthy district heating systems control
Milan Zdravkovic
Data scientist and associated professor @ Faculty of Mechanical Engineering, University of Nis
Once you step aside from the glorified landscape of generative AI with its perceived hyper-realism and hypnotizing illusion of creativity, you find yourself in the world of automation, THE actual billion-dollar holy grail of the artificial intelligence. In this talk, I will show how to use AI to solve industrial control automation problem by relying on the forecasts of heat demand in residential areas. How Explainable AI can make this solution trustful? How to interpret different kinds of feature importances? What are the anchors, prototypes and counterfactual explanations and how we can use those to deconstruct empirical, data-driven into formal knowledge which can be understood and validated by district heating experts?
12:30 – 12:45 ET: Observing the Future
Aleksandar Lukic
Machine Learning Engineer @ Planet Soft
ET doesn’t stand for Extra-Terrestrial, but rather for Emerging Technology, Efficiency Toolkit, Enhanced Telemetry, and Enterprise Transformation. This system represents a new era in AI-driven solutions for energy infrastructure, specifically designed to transform how high-voltage power lines are inspected and maintained. ET leverages drone-collected LIDAR, RGB, and thermal data, processed through advanced machine learning and data analytics, to enable utility companies to proactively identify and address potential issues. This continuous, intelligent analysis allows for the precise detection of malfunctions and anomalies, reinforcing the stability and security of Europe’s energy grid. ET is more than a system—it’s a vision of a resilient and forward-thinking energy future.
13:00 – 14:00 Panel duscussion
Moderator: Lisa Kratochwill
Startup Engagement Manager @ VERBUND AG
12:00
–
14:00
Data-Driven Energy Management
Francesco Conti & Milan Zdravkovic & Aleksandar Lukic & Lisa Kratochwill
Data Scientist @ Agile Lab & Data scientist and associated professor @ Faculty of Mechanical Engineering, University of Nis & Machine Learning Engineer @ Planet Soft & Startup Engagement Manager @ VERBUND AG
Edge AI
Breakout Session
Description: Edge AI is rapidly transforming the way AI systems operate, shifting intelligence from centralized clouds to the edges of networks. This session begins by exploring the evolution from basic Edge AI to Edge Intelligence, Awareness, and even Consciousness, introducing the concept of autonomous and “thinking” edge devices. We will then dive into the technical aspects of optimizing large language models (LLMs) for on-device deployment, examining how model compression techniques can enhance performance while reducing costs. Finally, we’ll look at the real-world application of smart agents in dynamic environments, highlighting the challenges and innovative solutions that make scalable, autonomous edge AI possible. Join us to explore the future of edge computing, where AI becomes truly distributed and intelligent at the edge.
15:00 – 15:15 From Edge AI to Edge Intelligence, Awareness and Consciousness
Salvatore Distefano
Professor @ University of Messina
This talk proposes a journey into Edge AI and Edge Intelligence, towards “thinking Edges”. Edge AI refers to the deployment of AI algorithms on edge devices, enabling local data processing with benefits like reduced latency and enhanced privacy. Edge Intelligence broadens this concept to data collection, caching, processing and analysis (training and inference), as well as task offloading, close to the data source. Moving a step forward, the application of Edge Intelligence to the edge device management brings to autonomous edges able to self-manage, self-configure and self-healing. In this roadmap, a key element is Edge Awareness, introducing cognitive capabilities on edges to enable intelligent device monitoring for identifying potential issues and related causes. Then, Edge Consciousness allows to predict events on edges and proactively react to them. This involves leveraging predictive models for future event forecasting and prescriptive analytics for decision-making support and automation. Thereby, starting from Edge AI and combining Edge Intelligence, Awareness, and Consciousness, edge devices can achieve a higher level of autonomy, enabling intelligent decision-making, proactive problem-solving, and enhanced efficiency in several application domains, i.e. “thinking Edges”.
15:30 – 15:45 Smart Agents for real-world edge Applications
Miguel de Prado
Programme Manager @ VERSES AI
This talk will explore the challenges and limitations of current deep learning systems, focusing on the critical barriers to deploying AI in real-world edge applications. We will examine solutions to overcome these obstacles, enabling more efficient, scalable AI at the edge. Additionally, the talk will highlight cutting-edge trends and techniques that facilitate continuous learning and adaptation of models in dynamic and changing environments, empowering smart edge agents to operate autonomously and in complex settings.
15:45 – 16:00 TBA
Milovan Medojevic
Senior researcher, research group lead @ The Institute for Artificial Intelligence Research and Development of Serbia
TBA
16:00 – 17:00 Panel duscussion
Moderator: Radovan Bacovic
Staff Data Engineer @ GitLab
15:00 – 15:15 From Edge AI to Edge Intelligence, Awareness and Consciousness
Salvatore Distefano
Professor @ University of Messina
This talk proposes a journey into Edge AI and Edge Intelligence, towards “thinking Edges”. Edge AI refers to the deployment of AI algorithms on edge devices, enabling local data processing with benefits like reduced latency and enhanced privacy. Edge Intelligence broadens this concept to data collection, caching, processing and analysis (training and inference), as well as task offloading, close to the data source. Moving a step forward, the application of Edge Intelligence to the edge device management brings to autonomous edges able to self-manage, self-configure and self-healing. In this roadmap, a key element is Edge Awareness, introducing cognitive capabilities on edges to enable intelligent device monitoring for identifying potential issues and related causes. Then, Edge Consciousness allows to predict events on edges and proactively react to them. This involves leveraging predictive models for future event forecasting and prescriptive analytics for decision-making support and automation. Thereby, starting from Edge AI and combining Edge Intelligence, Awareness, and Consciousness, edge devices can achieve a higher level of autonomy, enabling intelligent decision-making, proactive problem-solving, and enhanced efficiency in several application domains, i.e. “thinking Edges”.
15:30 – 15:45 Smart Agents for real-world edge Applications
Miguel de Prado
Programme Manager @ VERSES AI
This talk will explore the challenges and limitations of current deep learning systems, focusing on the critical barriers to deploying AI in real-world edge applications. We will examine solutions to overcome these obstacles, enabling more efficient, scalable AI at the edge. Additionally, the talk will highlight cutting-edge trends and techniques that facilitate continuous learning and adaptation of models in dynamic and changing environments, empowering smart edge agents to operate autonomously and in complex settings.
15:45 – 16:00 TBA
Milovan Medojevic
Senior researcher, research group lead @ The Institute for Artificial Intelligence Research and Development of Serbia
TBA
16:00 – 17:00 Panel duscussion
Moderator: Radovan Bacovic
Staff Data Engineer @ GitLab
15:00
–
17:00
Edge AI
Salvatore Distefano & Miguel de Prado & Milovan Medojevic & Radovan Bacovic
Professor @ University of Messina & Programme Manager @ VERSES AI & Senior researcher, research group lead @ The Institute for Artificial Intelligence Research and Development of Serbia & Staff Data Engineer @ GitLab
Elytis
Data-Driven Mobility
Breakout Session
Description: Deep learning technologies are revolutionizing the mobility sector, providing innovative solutions for optimizing transportation routes, predicting demand, and enhancing vehicle performance. By harnessing large datasets and cutting-edge algorithms, companies are able to improve operational efficiency, reduce costs, and elevate the overall mobility experience. This session explores real-world applications such as predictive maintenance for electric vehicles, smart infrastructure integration, and real-time traffic management. Discover how deep learning is reshaping the future of mobility, driving innovation across industries like automotive manufacturing and rail transport, and streamlining complex logistics systems.
10:00 – 10:15 AI Solutions for Mission-Critical Railway Applications
Nenad Mijatovic
Director of Data Science and AI @ Alstom
This presentation explores two innovative AI use cases in the mobility industry. The first AI solution for Communication-Based Train Control (CBTC)—an essential component for safe train operations—detects radio communication failures that can disrupt services and affect passenger satisfaction. With over ten projects implemented, we have significantly reduced on-site inspections and emergency brake incidents, resulting in smoother and more reliable train operations. The second solution, the smart troubleshooting system, combines predictive maintenance expertise with human-centric AI technologies, including a Large Language Model (LLM). This empowers engineers and maintenance teams to quickly identify the causes of failures and minimize operational costs through data quality checks and causal analysis. These examples showcase Alstom’s leadership in AI integration, highlighting our commitment to creating trustworthy solutions that enhance system performance in the mobility sector. Join us to explore these advancements in AI technology.
10:15 – 10:30 Data Science for Future of Mobility
Dordije Tripkovic
Product Lead – Data Science for Batteries @ BASF
Electric vehicles require durable (long life), high capacity (long range), and affordable batteries to be competitive with internal combustion engine cars. A crucial component which determines the overall battery performance is cathode and more precisely cathode active material. BASF is one of the largest producers of cathode active materials with a global footprint and partnerships with various car manufacturers. To satisfy the customer requirements timely as well as to guide the development of novel battery materials, Battery Materials R&D was one of the pioneers in the company to introduce data science methods on large scale. This talk will highlight some of the use cases and the practical implications for the business.
10:30 – 10:45 Foursquare’s Movements Engine: From Raw Data to Normalized User Behavior
Stefan Hacko
Senior Data Scientist @Foursquare
One might be quick to assume that if we have access to the user’s GPS location data, we can easily determine the user’s behavior during a day (e.g. what places did a user visit). It is easy to overlook that the user’s data is usually noisy, it often has gaps due to loss of GPS signal, and without additional context raw GPS pings are hard to interpret. On top of that processing geospatial data at petabyte scale requires significant computational resources, streamlined pipelines, and maximally optimized algorithms. In this talk we present how Foursquare built a robust movements engine that processes raw user location pings into actionable insights at scale. We discuss the methodologies for adding context to raw data to derive meaningful stops and visits, addressing the major complexities of spatial-temporal analysis. Our engine leverages advanced machine learning algorithms to filter, cluster, and interpret location data, ensuring accuracy and privacy while minimizing biases. Finally, we explore the normalization techniques used to correct for the inherent spatial-temporal biases in the data, providing a reliable foundation for downstream analytics and decision-making.
10:45 – 11:00 TBA
Marko Vasiljevski
Global Head of Data Science @ JATO Dynamics Ltd
TBA
11:00 – 12:00 Panel duscussion
Moderator: Marko Vasiljevski
Global Head of Data Science @ JATO Dynamics Ltd
10:00 – 10:15 AI Solutions for Mission-Critical Railway Applications
Nenad Mijatovic
Director of Data Science and AI @ Alstom
This presentation explores two innovative AI use cases in the mobility industry. The first AI solution for Communication-Based Train Control (CBTC)—an essential component for safe train operations—detects radio communication failures that can disrupt services and affect passenger satisfaction. With over ten projects implemented, we have significantly reduced on-site inspections and emergency brake incidents, resulting in smoother and more reliable train operations. The second solution, the smart troubleshooting system, combines predictive maintenance expertise with human-centric AI technologies, including a Large Language Model (LLM). This empowers engineers and maintenance teams to quickly identify the causes of failures and minimize operational costs through data quality checks and causal analysis. These examples showcase Alstom’s leadership in AI integration, highlighting our commitment to creating trustworthy solutions that enhance system performance in the mobility sector. Join us to explore these advancements in AI technology.
10:15 – 10:30 Data Science for Future of Mobility
Dordije Tripkovic
Product Lead – Data Science for Batteries @ BASF
Electric vehicles require durable (long life), high capacity (long range), and affordable batteries to be competitive with internal combustion engine cars. A crucial component which determines the overall battery performance is cathode and more precisely cathode active material. BASF is one of the largest producers of cathode active materials with a global footprint and partnerships with various car manufacturers. To satisfy the customer requirements timely as well as to guide the development of novel battery materials, Battery Materials R&D was one of the pioneers in the company to introduce data science methods on large scale. This talk will highlight some of the use cases and the practical implications for the business.
10:30 – 10:45 Foursquare’s Movements Engine: From Raw Data to Normalized User Behavior
Stefan Hacko
Senior Data Scientist @Foursquare
One might be quick to assume that if we have access to the user’s GPS location data, we can easily determine the user’s behavior during a day (e.g. what places did a user visit). It is easy to overlook that the user’s data is usually noisy, it often has gaps due to loss of GPS signal, and without additional context raw GPS pings are hard to interpret. On top of that processing geospatial data at petabyte scale requires significant computational resources, streamlined pipelines, and maximally optimized algorithms. In this talk we present how Foursquare built a robust movements engine that processes raw user location pings into actionable insights at scale. We discuss the methodologies for adding context to raw data to derive meaningful stops and visits, addressing the major complexities of spatial-temporal analysis. Our engine leverages advanced machine learning algorithms to filter, cluster, and interpret location data, ensuring accuracy and privacy while minimizing biases. Finally, we explore the normalization techniques used to correct for the inherent spatial-temporal biases in the data, providing a reliable foundation for downstream analytics and decision-making.
10:45 – 11:00 TBA
Marko Vasiljevski
Global Head of Data Science @ JATO Dynamics Ltd
TBA
11:00 – 12:00 Panel duscussion
Moderator: Marko Vasiljevski
Global Head of Data Science @ JATO Dynamics Ltd
10:00
–
12:00
Data-Driven Mobility
Nenad Mijatovic & Dordije Tripkovic & Stefan Hacko & Marko Vasiljevski
Director of Data Science and AI @ Alstom & Product Lead – Data Science for Batteries @ BASF & Senior Data Scientist @Foursquare & Global Head of Data Science @ JATO Dynamics Ltd
Socrates
Workshop: Lab based on Microsoft Certified: Azure AI Engineer Associate Microsoft Learning materials and labs
Workshop
This workshop offers hands-on experience with Azure AI Vision and Speech capabilities. Participants will learn to analyze images using pre-built models for generating captions, detecting objects and people, and performing background removal. Additionally, the workshop covers Azure AI Speech translation, enabling attendees to translate spoken language and synthesize the output into spoken responses—empowering practical AI solutions for real-world scenarios.
11:00
–
12:00
Workshop: Lab based on Microsoft Certified: Azure AI Engineer Associate Microsoft Learning materials and labs
Catalin Gheorghiu
Solutions Architect @ EPAM
* this program is not final and is subject to change, full schedule will be available soon