From Vibe-coded PoC to Production: A Healthcare Master Data Management System

Our first step in the vibe code cleanup was to establish a more consistent and reliable application architecture. Since the original codebase was overwhelmed with duplication, tight coupling, and difficulties maintaining existing functionality, we reorganized the code into well-defined layers and functional slices, improving separation of concerns and making individual components easier to understand and maintain. We also addressed duplicated and dead code where relevant and repaired the logic behind malfunctioning features. These changes established a clearer structure for the integrated modules and created an architectural baseline the client can use to improve the remaining application.

Next, our experts reviewed and corrected the logic affecting the planned integration, resolving issues that could cause crashes or prevent features from working correctly. We also implemented missing features required to connect the client’s core project with the MDM system. Because some of the affected code was shared across the application, these improvements also benefited other areas that depended on the same components.

Our audit revealed inconsistencies between frontend and backend role-based access control that we needed to address. While the interface reflected user permissions, backend authorization did not consistently enforce the same rules, creating a potential security gap. We fixed it by aligning the frontend and backend permission models and strengthening backend role validation to ensure that access restrictions were enforced consistently. This helped to establish a reliable access-control foundation for the app.

The existing code structure and query patterns made data operations less efficient and caused performance issues. We reviewed and optimized the queries used by the affected modules, improving data access and reducing unnecessary processing. These changes helped the integrated functionality operate more efficiently while supporting a more maintainable approach to database interactions.

To build a dependable way for system components to communicate without becoming tightly coupled, we introduced Apache Kafka. This event-streaming platform provided a more structured approach to exchanging events and supported the integration requirements without adding further direct dependencies between components. While the existing tech stack remained in place, Apache Kafka only extended it with a messaging capability needed for the new functionality and future integration scenarios.

We improved code separation and expanded unit and integration testing for the refactored modules. Our QA specialists tested the updated functionality and developed testing strategies to support reliable validation as the application evolves. We also added monitoring and documentation to simplify maintenance and provide a blueprint for improving the rest of the application. By preserving the existing CI/CD pipelines and Dev, UAT, and production environments, we focused on addressing the key engineering gaps.