Understanding the Design of Edge AI-based Systems

Nome do aluno Jander Pereira Santos Junior
Título do trabalho Understanding the Design of Edge AI-based Systems
Resumo do trabalho The rapid convergence of Artificial Intelligence (AI) and Edge Computing has accelerated the emergence of Edge AI systems, in which intelligent capabilities are deployed across the device--edge--cloud continuum under heterogeneous, resource-constrained, and dynamic conditions. Despite growing adoption, the architectural foundations of these systems remain fragmented and under-documented, limiting our ability to understand how such architectures are structured and justified in practice.
We conducted a comprehensive empirical study grounded in systematic repository mining and evidence-based analysis of open-source projects, based on 400 architectural fragments extracted from 57 high-activity repositories. To structure extraction, coding, and interpretation, we applied a multi-level coding scheme explicitly aligned with ISO/IEC/IEEE 42010 (architecture description evidence), ISO/IEC 25010 (quality requirements), and ISO/IEC 30141 (IoT reference capabilities), enabling the identification of recurring patterns, design priorities, and documentation practices.
From this corpus, we derived an architectural axis by levels of abstraction: Meta-Architecture, Platform/Infrastructure, System, and Subsystem, and characterized recurring capabilities and quality concerns observable in repositories.
Finally, we synthesized a catalog of 52 practitioner-validated architectural guidelines, formally aligned with the adopted ISO taxonomies. Together, our results provide an empirically grounded blueprint of open-source Edge AI architectural practice, supporting both engineering decision-making and future research.
Orientador Eduardo Almeida
Membro Titular 1 (com afiliação) Nelio Alessandro Azevedo Cacho (UFRN)
Link para o curriculum lattes https://lattes.cnpq.br/4635320220484649
Membro Titular 2 (com afiliação) Rita Suzana Pitangueira Maciel (IC-UFBA)
Link para o curriculum lattes https://lattes.cnpq.br/0290950513716961
Suplente 1 (com afiliação) Frederico Araujo Durao (IC-UFBA))
Suplente 2 (com afiliação) Elisa Yumi Nakagawa (ICMC/USP)
Data do exame 09 Apr, 2026
Horário do exame 2:00 PM

 

Data da Defesa: 
09/04/2026 - 14:00
Tipo de Defesa: 
Qualificação de Mestrado