Workshop opening followed by Pitch Session 1 (5-minute pitch of each paper)
Pitch Session 2 (5-minute pitch of each paper) followed by breakout group discussions
Keynote: Henry Muccini (University of L’Aquila, Italy) Title: Green AI: From Industry Apathy to Architectural Opportunity
Abstract: As Artificial Intelligence (AI) becomes a foundational layer in modern software, the discipline of Green Software Engineering must expand its focus from model optimization to systemic architectural efficiency. This presentation discusses a three-tiered hierarchy of AI sustainability, characterizing the transition from model-bound energy consumption to the emergent, architecture-driven overhead of LLM agents. We begin by examining the broad landscape of Green AI, distinguishing between the inherent energy demands of AI components and the external role of AI as a tool for reducing the energy consumption of traditional software applications. We then transition to a technical analysis of Large Language Model (LLM) inference, where energy expenditure remains largely decoupled from software architecture. The core of our analysis addresses the emergent paradigm of Agentic AI systems. We argue that at this level, energy consumption becomes a highly architectural concern. Unlike standard LLMs, agents rely on iterative reasoning loops and tool-calling mechanisms that necessitate a “Shift Left” approach in the design phase. We investigate, among other factors, the impact of token management and the strategic deployment of Small Language Models. Finally, we present preliminary empirical evidence from a multi-faceted study—comprising a Systematic Literature Review (SLR), practitioner interviews, and an analysis of open-source repositories. Our findings reveal a significant “awareness gap”: while the technical complexity of agents grows, industry practitioners currently prioritize speed and cost savings over sustainability. We conclude by identifying research opportunities and establishing energy efficiency as a first-class citizen in autonomous agent design.
Short Bio: I am a Professor of Software Engineering at the University of L’Aquila, Italy, where I lead the FrAmeLab research laboratory, part of the SWEN software engineering research group. My work operates at the critical intersection of software architecture, AI engineering, and sustainable ICT. With over two decades of contributions to the software architecture community, my current research focuses on the systemic architectural efficiency of AI-driven systems. I am specifically investigating the transition from model-centric energy consumption to the complex, architecture-driven overhead inherent in LLM-based agentic systems. Through empirical methods—including systematic literature reviews and repository mining—his team is defining the “Shift Left” patterns required to bridge the current industry awareness gap in AI sustainability. My commitment to the research community is reflected in my service as Chair of the Steering Committee for the IEEE International Conference on Software Architecture (ICSA) and as a member of the Steering Committee for the European Conference on Software Architecture (ECSA) and the Conference on AI Engineering (CAIN). As Associate Editor-in-Chief of IEEE Software, I provide my contribution to the peer-review and dissemination of research that may have an impact on industrial practices. Since 2019, my work has focused on bridging the gap among software architecture, machine learning, and sustainability through participation in international workshops, Dagstuhl seminars, and specialized publications. I earned my Ph.D. in Computer Science from the University of Rome “La Sapienza,” followed by a postdoctoral fellowship at the University of California, Irvine
Breakout group discussions (continuation), presentation of results, and closing session
Social event