The increasing prevalence of mental illnesses, including stress-related disorders, has made workplace strain a growing concern, particularly for knowledge workers who face high cognitive demands, non-linear workflows and tight deadlines. The complex interplay between workplace conditions, self-management strategies, and individual health remains largely under-studied, especially in real-world settings, and help-seeking behavior to mitigate health and well-being challenges remains low. Although advances in digitalization facilitate the capture and analysis of workplace-specific contexts, they have not yet been fully leveraged in multimodal diagnosis and intervention systems for pattern recognition that may enable targeted interventions, leaving a gap for a comprehensive solution. To address this gap, we applied the Design Science Research approach to develop an architecture for a multimodal diagnosis and intervention platform. Using a mixed-methods study, we derived requirements from a comprehensive literature review, semi-structured interviews (n = 12), and a survey (n = 32) with knowledge workers. The derived design principles guided the development of a process-oriented architecture, which was evaluated and refined based on expert feedback (n = 6). Our work provides a basis for future research and practical implementations to advance workplace well-being technologies.