DevFinder: an architecture for recommending software development specialists with industry-specific expertise in desired technologies

Abstract

Global Software Development (GSD) has experienced a significant surge in recent years, driven by the growing importance of software across various industries. The demand for experts with specific technical skills and industry domain experience has become crucial for organizations to maintain a competitive edge. However, the rapid pace of change characteristic of the fourth industrial revolution presents significant challenges in meeting this demand. This revolution evolves exponentially, a ecting all countries and industries simultaneously, and necessitates the constant adaptation of the software industry to keep up with swift advancements. To address the talent gap and the challenges posed by the rapid evolution of technology, this study introduces DevFinder, a software architecture designed to support the search for software development specialists whose knowledge aligns with specific industry domains and technologies. DevFinder leverages advanced techniques such as complex networks and Large Language Models (LLMs) to process information from real-world code repositories, generating a ranked list of developers who meet stakeholders’ specific needs. By extracting, interpreting, and processing data from software repositories, DevFinder aims to bridge the gap between the technical requirements of companies and the scarcity of specialized talent. The implementation of DevFinder architecture was carried out through three cycles of Design Science Research, culminating in the current f inal version. The architectural requirements were identified through a systematic mapping study, and the implementation iterations were assessed in three separate studies. The evaluation results confirm that the developed architecture is capable of supporting job recruiters in finding software specialists with the desired combination of industry domain knowledge and technical expertise. The main contributions of this research include specifying the requirements for an e ective architecture, developing a business process model that communicates the extraction, filtering, and ranking processes to stakeholders, and implementing processes to extract and filter repositories using LLM techniques. Additionally, a metric was defined to rank committers based on collaboration aspects, and the proposed architecture was implemented with real-world data from repositories and software developers. Overall, DevFinder demonstrates the potential to aid organizations in locating suitable software developers, thereby addressing the talent gap and ensuring that software development companies can keep pace with the fourth industrial revolution’s rapid advancements