Dr. Sam Leewis

HU University of Applied Sciences Utrecht, Netherlands

Dr. Sam Leewis is a researcher and lecturer at Utrecht University of Applied Sciences, where he works within the Digital Ethics research group. His research focuses on decision mining, responsible digital innovation, digital twins, and the use of data and AI in public-sector organizations. He is particularly interested in how digital technologies can improve operational decision-making while safeguarding public values such as transparency, accountability, and equality. His work combines applied research, education, and collaboration with governmental organizations and industry partners.

Title: Can we use generative AI to uncover the business logic hidden inside complex software systems?

Abstract: This talk explores how Large Language Models (LLM) can help reverse-engineer source code into understandable and standardized decision models. Legacy systems often contain decades of accumulated business knowledge, while documentation is incomplete, outdated, or simply missing. Replacing these systems can therefore mean losing essential organizational knowledge. Our research investigates whether LLMs can bridge this gap by automatically identifying decision logic in source code and translating it into Decision Model and Notation (DMN). Based on the development and evaluation of this prototype, the talk proposes an end-to-end approach in which LLMs identify relevant decisions, reconstruct decision dependencies and rules, and generate visual DMN models. Beyond presenting what works, the talk also discusses where LLMs still struggle and what this means for using generative AI to understand, maintain, and ultimately modernize legacy software.

Assoc. Prof. Anibal Tavares de Azevedo

State University of Campinas (UNICAMP), Brazil

Anibal Tavares de Azevedo is a highly accomplished Associate Professor ("Livre-Docente") at the University of Campinas (UNICAMP). His research lies at the intersection of Operations Research, Combinatorial Optimization, System Simulation, and Artificial Intelligence applied to large-scale logistics, energy, and healthcare systems.

He is the founder, chief architect, and coordinator of the BISS (Big Interconnected Smart Science) Project and the R3PO (Rede Paulista de Pesquisa Operacional) Network. Through these pioneering initiatives, Prof. Azevedo has established a groundbreaking framework for decentralized, citizen-driven distributed science, mobilizing hundreds of researchers to solve high-impact, real-world public logistics challenges.

Title: The Green Compute Synergy: Sovereign AI, Volunteer Networks, and the Brazil-Japan Energy-Technology Win-Win

Abstract: The Global AI Infrastructure Crisis: The presentation opens by addressing the severe physical and electrical bottlenecks facing centralized, hyperscale data centers (such as multi-year grid interconnection queues and community noise complaints).

The NVIDIA XFRA Residential Edge Solution: We introduce Span's UL 3141-certified smart electrical panels and XFRA's liquid-cooled edge compute nodes (packing 16 NVIDIA RTX Pro 6000 Blackwell GPUs), demonstrating how they leverage existing residential electrical headroom to run high-performance AI inference close to users.

Brazil's Renewable Energy Paradox (The "Duck Curve"): We examine the contemporary energy challenge in Brazil, where a massive 50GW surplus of distributed solar generation causes overproduction emergencies and forced ONS grid curtailments at noon, while fossil-fuel plants are still activated during night-time spikes.

The Brazil-Japan Strategic Alliance: We propose a global win-win. Brazil provides abundant, clean, and low-cost solar surplus to power edge XFRA nodes during peak solar hours, solving its grid over-generation crisis. Japan and NVIDIA provide the advanced hardware and open-source software—specifically NVIDIA Nemotron Open Models—enabling highly secure, localized, and sovereign AI applications.

The BISS Methodology and the R3PO Volunteer Network: We explain our operational framework. The BISS (Big Interconnected Smart Science) platform fragments complex logistics simulations (like emergency ambulance location in Campinas and São Paulo) into reproducible, cryptographic "Work Units" via unique Combination IDs and MD5 hashing, validated by a factor-2 Biss Checker redundancy check. These tasks are solved by the R3PO student volunteer network and tracked on a gamified, real-time Status Map.

Cognitive Cities and the DeSci Economy: We conclude with the "Sense-Think-Act" loop of edge cities where localized Nemotron models safely optimize urban systems, tokenizing validated resource efficiency into IP-NFTs to fund public university research.