BEST SOFTWARE ARCHITECTURES FOR SYSTEMS WITH ARTIFICIAL INTELLIGENCE
DOI:
https://doi.org/10.56238/ramv20n16-018Keywords:
Artificial Intelligence, Software Architecture, Microservices, MLOps, RAGAbstract
The rapid advancement of Artificial Intelligence (AI) has demanded that software engineering teams adopt robust, scalable, and observable architectures to support machine learning models in production. This work presents a comparative analysis of the main software architectures currently used in the development and deployment of AI-based systems, addressing patterns such as microservices, serverless, MLOps pipelines, event-driven architecture, and RAG (Retrieval-Augmented Generation) systems. The trade-offs of each approach, selection criteria, and emerging trends for 2025 and beyond are discussed.
References
BERKELEY AI RESEARCH INSTITUTE. The shift from models to compound AI systems. Disponível em: <https://bair.berkeley.edu>. Acesso em: 27 maio 2026.
KREUZBERGER, Dominik; KÜHL, Niklas; HIRSCHL, Sebastian. Machine learning operations (MLOps): overview, definition, and architecture. IEEE Access, v. 11, p. 31866-31879, 2023.
LAKATOS, Eva Maria; MARCONI, Marina de Andrade. Fundamentos de metodologia científica. 3. ed. rev. e ampl. São Paulo: Atlas, 1991. 270 p.
LEWIS, Patrick et al. Retrieval-augmented generation for knowledge-intensive NLP tasks. In: ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS, 33., 2020, [S. l.]. Anais [...]. [S. l.]: NeurIPS, 2020. p. 9459-9474.
SCULLEY, D. et al. Hidden technical debt in machine learning systems. In: ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS, 28., 2015. Anais [...]. [S. l.]: NeurIPS, 2015.
SUTHERLAND, Jeff. Scrum: a arte de fazer o dobro do trabalho na metade do tempo. São Paulo: LeYa, 2014.
ZAHARIA, Matei et al. Accelerating the machine learning lifecycle with MLflow. IEEE Data Engineering Bulletin, v. 41, n. 4, p. 39-45, 2018.
