Advanced Software Architectures for Designing Human-Nudging Systems

Tiziano Santilli

Nudging systems steer human behavior toward better choices without removing individual freedom. This talk examines the architectural patterns and design decisions behind AI-driven nudging systems, focusing on how modularity, adaptability, and transparency determine whether interventions stay ethical, trustworthy, and effective. It links architectural quality attributes such as maintainability, explainability, and scalability to real behavioral outcomes, and explores the trade-offs designers face when technology must influence people responsibly. Examples from domains like energy use and mobility show how choices made at the architecture level translate into measurable sustainability impact. The core argument: treating human-nudging as an engineering discipline, grounded in sound software architecture, is essential for scaling behavioral technologies people can trust.

Fitting powerful AI models for low-power devices

Francesco Daghero

State-of-the-art AI models are powerful but often too hungry in memory, computation, and energy to run on the microcontrollers and embedded devices at the edge. Closing that gap is a software engineering challenge as much as a modeling one. This talk explores the techniques that make it possible to fit powerful models onto constrained hardware, from quantization and pruning to neural architecture search and hardware-aware compilation. It looks at the trade-offs among accuracy, latency, and power that shape every edge deployment, and at how automated tooling and AI compilers help navigate them. The broader focus is how on-device intelligence can be engineered to bring AI that is responsive, private, and energy-efficient beyond the cloud.

Cybersecurity Vulnerabilities and Opportunities of AI for Organizations and the Public

Sani Abdullahi and Maximilian von Zastrow

The rapid adoption of AI has created new opportunities, but also new security risks for organizations and public institutions. This talk explores the evolving cybersecurity landscape, examining both the challenges AI introduces and the ways it can strengthen cyber defenses.

Sani Abdullahi will discuss today's key cybersecurity challenges and how AI can be used to improve resilience against emerging threats. Maximilian von Zastrow will then present a real-world law enforcement case, demonstrating how AI-assisted password recovery can support digital investigations while highlighting the practical opportunities and considerations of applying AI in digital forensics.