Integrating Machine Learning with Software Development Lifecycles: Insights from Experts

This paper examines the challenges related to integrating machine learning (ML) development with software development lifecycle (SDLC) models. Data-intensive development and use of ML are gaining popularity in information systems development (ISD). To date, there is little empirical research that explores the challenges that ISD practitioners encounter when integrating ML development with SDLC frameworks. In this work we conducted a series of expert interviews where we asked the informants to reflect upon how four different archetypal SDLC models support ML development. Three high level trends in ML systems development emerged from the analysis, namely, (1) redefining the prescribed roles and responsibilities within development work; (2) the SDLC as a frame for creating a shared understanding and commitment by management, customers, and software development teams: and (3) method tailoring. This study advances the body of knowledge on the integration of conceptual SDLC models and ML engineering. Post-print / Final draft