Continuous Software Engineering Practices in AI/ML Development

This technical report investigates the challenges associated with the adoption of continuous software engineering (SE) practices in the development of machine learning (ML) systems. The study is based on a multiple case study approach involving thematic interviews with eight ML experts from various organizations. The aim is to explore continuous SE beyond the specific practices associated with MLOps, which is a narrower focus within the broader continuous SE framework. The report identifies and discusses the various challenges organizations face when attempting to implement all 13 continuous SE practices as outlined in existing literature. Key findings reveal that communication issues are a significant barrier, as ML experts often work in silos, disconnected from other project stakeholders and customers. The report emphasizes the need for improved collaboration across different roles involved in ML development to enhance the integration of continuous SE practices.