DevOps and continuous delivery have impacted the organizational structures of development and infrastructure groups in software-producing organizations. Our research aims at revealing the different options adopted by the software industry to organize such groups, understanding why different organizations adopt distinct structures, and discovering how organizations handle the drawbacks of each structure. We interviewed 68 carefully-selected IT professionals, 45 working in Brazil, 10 in the USA, 8 in Europe, 1 in Canada, and 4 in globally distributed teams. By analyzing these conversations through a Grounded Theory process, we identified conditions, causes, reasons to avoid, consequences, and contingencies related to each discovered structure (segregated departments, collaborative departments, API-mediated departments, and single department). In this way, we offer a theory to explain organizational structures for development and infrastructure professionals. This theory can support practitioners and researchers in comprehending and discussing the DevOps phenomenon and its related issues, and also provides valuable input to practitioners’ decision-making. Published as: L. Leite, N. Lago, C. Melo, F. Kon, and P. Meirelles, ”A theory of organizational structures for development and infrastructure professionals,” in IEEE Transactions on Software Engineering, vol. 49, no. 4, pp. 1898-1911, 1 April 2023, doi: 10.1109/TSE.2022.3199169.

Machine Learning (ML) has grown in popularity in the software industry due to its ability to solve complex problems. Developing ML Systems involves more uncertainty and risk because it requires identifying a business opportunity and managing the source code, data, and trained model. Our research aims to identify the existing practices used in the industry for building ML applications. The goal is to comprehend the orga- nizational complexity of adopting ML Systems. We conducted a Multivocal Literature Review (MLR) and used Grounded Theory (GT) to build a taxonomy with the practices applied to the ML System lifecycle from the industry and academic perspectives. We selected 41 posts from grey literature and 37 papers from scientific repositories. Following a systematic GT protocol, we mapped 91 practices, grouped in 6 core categories related to designing, developing, testing, and deploying ML Systems. The results can help organizations identify the gaps in their current ML processes and practices, and provide a roadmap for improving and optimizing their ML systems. The comprehensive taxonomy of practices developed in this research serves as a valuable tool for managers, practitioners, and researchers in the ML field, providing a clear and organized understanding of the complexity of managing ML systems.

Frameworks are essential for software development, providing code design and reuse for its users. Well-known Java frameworks and APIs such as Spring, JUnit, and JPA rely on metadata usage, commonly defined by code annotations. Those frameworks use the Java Reflection API for consuming and processing these annotations. Code elements usually have some similarities and can also have the same annotation. This paper proposes a model for defining conventions over configuration for annotations usage decoupled from the metadata reading logic. With this model, if a convention is present, the framework will consider that the element is configured by a specific target annotation, even if the code element does not have the annotation. We implemented this model in the metadata reading framework Esfinge Metadata API. The model implementation was evaluated refactoring an existing framework to add support to conventions using our approach. As a result, it was possible to introduce the conventions only by adding configurations to the annotations. The model was further evaluated in an experiment in which participants implemented the Conventions over Configuration pattern using the Esfinge Metadata API and Java Reflection API. Based on the results, approach fulfilled its goal of supporting the definition of conventions decupled from the framework logic, making the code more readable and easier to maintain according to the participants perception.