Forming effective teams is an essential but challenging task, especially for organizations that carry out multiple projects simultaneously, a problem known as the Multiple Team Formation (MTF) problem. The literature presents several solutions for the MTF problem, mostly modeling it as a search problem. However, the existing solutions are not suitable for Scrum projects. We addressed this gap by developing an approach composed of two main steps. First, we designed a Structured Task Model to support creating developers' profiles given their performance on past Scrum projects. Then, given a set of target projects' technology requirements and the available developers' profiles, we developed a Genetic Algorithm to form the teams for a set of target projects. We evaluated the proposed approach by comparing the teams formed by our approach with the ones formed by project managers from one organization. Our approach achieved 85% of precision when compared with the teams provided by the project managers who worked on the same target projects. We also recorded an acceptance rate of up to 75%. The significant value of precision achieved suggests that our approach can provide teams close to the project managers' expectations. In addition, our Structured Task Model offers a promising way to build technical profiles semi-automatically for Scrum developers.

The software industry has experienced the integration of artificial intelligence capabilities into applications, facing new challenges regarding software development. Despite research and industry contributions providing lessons learned and best practices, no study proposed a reference process for developing this type of software, and practitioners still struggle to establish a working process. Through a Grounded Theory study involving practitioners with experience in machine learning (ML) projects, this paper presents an emerging theory of how ML-based systems are developed. The reported results comprise key elements of a reference development process with its respective phases and activities.

Context Software analytics approaches have supported managers in making informed decisions regarding several software engineering problems. Team formation is a challenging one, being investigated by the research community as new approaches and tools have been proposed. However, the existing studies do not appropriately address the aspects and procedures to be adopted in the development of tools to meet most scenarios, besides not providing a concrete solution useful from a practical perspective. Aims This study provides a framework to support the development of solutions that can help managers form software teams. Method We interviewed a key practitioner from a software organization and analyzed the collected data to understand how the team formation problem is currently handled, identifying underlying aspects and challenges faced by the organization. Results We presented an overview of the proposed framework and the results of a preliminary evaluation performed by integrating a prototype into an enterprise system. Conclusions Our results provide a concrete solution for the team formation problem that can be integrated not only into a project management tool, but also into software analytics tools.