Public health surveillance in Argentina is constrained by both vertical and horizontal fragmentation of health information systems. Existing platforms, such as the National Health Surveillance System (SNVS) and the Integrated Health Record (Historia de Salud Integrada, HSI), operate as loosely coupled systems with limited interoperability, often relying on manual data exchange. This paper proposes a cloud-native architecture for real-time public health vulnerability assessment that integrates heterogeneous data sources, including clinical records, epidemiological data, and geospatial social determinants of health. The main contribution is a stream-centric modular monolith architecture that extends Hexagonal Architecture with an internal Kappa-style streaming layer, enabling continuous processing of domain events and computation of contextual indicators. These indicators are reintegrated into the domain model, establishing a controlled feedback loop between analytical processing and vulnerability assessment. The architecture combines Command Query Responsibility Segregation (CQRS) with the Transactional Outbox pattern to ensure consistent and reliable propagation of state changes throughout the system while preserving a unified deployment model. The proposed approach enables near-real-time vulnerability assessment at both individual and population levels, supporting context-aware decision-making in fragmented public health environments. The design is evaluated through a scenario-based architectural validation grounded in representative public health workflows.