Speakers and Guests

Talk: The Cost of Trust: The Real Challenges of AI Reliability

Abstract: In the era of rapid artificial intelligence adoption, trust is the ultimate currency. Organizations are racing to integrate AI into their core operations, drawn by the promise of unprecedented efficiency and innovation. However, a critical realization is emerging across the industry: AI trust is not a default state; it is a heavily engineered, continuous, and often expensive investment.

This keynote, "The Cost of Trust: The Real Challenges of AI Reliability," peels back the layers of what it truly takes to build, scale, and deploy dependable AI systems in the real world. While the market focuses on the "magic" of generative models and predictive algorithms, the real operational battles are fought elsewhere. We will explore the hidden costs of mitigating hallucinations, preventing model drift, defending against adversarial attacks, and navigating algorithmic bias. In non-deterministic systems, a failure isn't just a system crash—it's a breach of logic, fairness, or privacy. When AI acts unpredictably, the cost is measured not just in wasted compute, but in severe reputational damage and the rapid erosion of user confidence.
Through real-world case studies and pragmatic frameworks, this session explores the fragile intersection of AI innovation and operational stability, offering leaders, developers, and stakeholders a realistic roadmap for navigating the complexities of AI reliability

Talk: Artificial Intelligence and Data-Driven Innovation: Current Trends and Future Directions

Abstract: This talk explores recent advances in Artificial Intelligence and their role in enabling data-driven innovation across multiple domains. It highlights how modern AI techniques, particularly machine learning and natural language processing, transform raw data into intelligent decision-support systems. The presentation discusses key application areas while addressing current challenges, including data quality, model bias, and ethical considerations. Finally, it outlines emerging trends and future directions, emphasizing the need for responsible, explainable, and human-centered AI systems.

Talk: Quantum: From the Ground Up to the Next Level of Intelligence.

Abstract: Step into the fascinating world of quantum computing—where bits become qubits and possibilities multiply. This lightning-fast session will introduce you to the core principles of quantum mechanics that power quantum computers, including superposition, entanglement, and interference. Discover how quantum computing differs from classical computing, why it matters, and how tech giants like IBM and Microsoft are making it accessible to everyone. Whether you're a curious learner or a future quantum developer, this session will spark your imagination and show you how to start exploring quantum today.

 Talk: Reliable AI and AI for Reliability in Critical Embedded Systems.

Abstract: In this talk, we present our recent contributions to the design of embedded Systems-on-Chip (SoCs) operating in critical and constrained environments. As these systems are increasingly deployed in harsh conditions, their design must go beyond conventional trade-offs between performance, power, and area to explicitly integrate reliability as a first-class constraint. In particular, we address the growing challenge of embedding artificial intelligence within safety-critical systems, where both the hardware platform and the AI models must be jointly designed for robustness.

Our research spans multiple application domains, including healthcare, aeronautics, aerospace, and automotive systems. We introduce advanced reliability-driven design methodologies based on error correcting codes, with a focus on LDPC (Low-Density Parity-Check) codes, and demonstrate their efficient implementation on reconfigurable platforms such as SRAM-based FPGAs as well as on ASICs. These approaches enable fault-resilient data processing pipelines suitable for high-throughput embedded applications.
Beyond communication and processing reliability, we investigate system-level robustness through real-world case studies. We analyze the observability and fault sensitivity of microprocessor-based systems under high-altitude radiation, and in the context of healthcare electronics, we develop optimized embedded architectures for accurate and energy-efficient EEG signal classification.
The second part of the talk focuses on the emerging paradigm of reliable embedded AI. We explore how to co-design AI algorithms and hardware architectures to ensure resilience against transient faults and error propagation. This includes studying the vulnerability of neural networks mapped onto reconfigurable circuits, and developing mitigation strategies at multiple levels from data representation and model compression to hardware-aware training and fault-tolerant architectures.
As a forward-looking perspective, we aim to leverage concepts from learning theory and bio-inspired computing to design inherently robust multicore and AI-driven embedded systems. The ultimate goal is to establish a new generation of embedded AI architectures that are not only efficient, but intrinsically reliable by design, meeting the stringent requirements of next-generation critical applications.