Darwoft developed an AI-powered document processing solution that extracts key medical information from unstructured records to accelerate clinical risk evaluations. The system integrates React, Node.js, Kotlin, and Kubernetes for scalability and compliance with healthcare standards. This innovation streamlines Risk Adjustment and Medicare Advantage processes, improving data accuracy and operational efficiency.
The Problem
The client required a scalable system capable of processing large quantities of unstructured medical documents while extracting relevant findings through a Language Model (LLM). These findings were key inputs for clinical risk evaluation processes, such as Medicare Advantage programs or Risk Adjustment Factor (RAF) calculations.
Existing methods relied heavily on manual review and coding, which slowed operations and limited data quality.
The Challenge
The project presented several technical and operational challenges, including:
Migrating the deployment model from ECS to EKS, while implementing a managed architecture using ArgoCD and the App of Apps pattern.
Developing an application for massive upload and automated processing of medical files through the LLM.
Ensuring traceability and version control for all model-generated results.
The challenge was to build an automated, explainable, and secure platform that integrated seamlessly into the client’s existing healthcare data ecosystem while maintaining HIPAA compliance and interoperability standards.
Integrating the solution with the client’s existing data ecosystem under strict healthcare regulations and interoperability protocols.
The Solution
Darwoft engineered a comprehensive end-to-end system combining a React and Node.js frontend with a Kotlin-based backend connected to a PostgreSQL database.
The backend centralized all business logic, exposing data through REST APIs while maintaining complete traceability of every processed document. The frontend provided an efficient, modern interface for uploading, reviewing, and supervising medical document processing.
To support scalability and maintainability, Darwoft redesigned the infrastructure deployment by decoupling Infrastructure as Code (IaC) components from the CI/CD pipeline. This allowed for independent management of infrastructure creation and image deployment in Kubernetes, maintaining backward compatibility with other organizational verticals still transitioning from ECS.
The architecture ensured a controlled and gradual migration, enabling the SRE team to manage infrastructure efficiently while ensuring stability in production environments.
Our AI-driven solution transformed the way clinical data is processed, turning unstructured medical documents into actionable insights — empowering healthcare organizations to evaluate risk faster, with greater accuracy and confidence.
The Conclusion
The implemented solution empowered the healthcare organization to automate the extraction of clinical data from unstructured documents, significantly reducing manual review and coding efforts.
Key benefits included:
Automated data extraction from unstructured sources, accelerating risk evaluation processes.
Integration with the corporate medical data pipeline, enhancing data quality and accessibility for analytics.
Scalable, manageable infrastructure, simplifying SRE operations and ensuring high availability.
Modern, user-friendly interface, improving user experience and process oversight.
By leveraging artificial intelligence and cloud-native architecture, Darwoft enabled the client to identify clinically relevant conditions directly from unstructured medical records, optimizing processes critical to Risk Adjustment and Medicare Advantage programs. The result: faster evaluations, more accurate condition recording, and stronger data-driven decision-making across the healthcare ecosystem.
Want to start a project?
Related blogs
If you found this Case Study insightful, you might also be interested in these related articles exploring similar topics in tech, design, and digital strategy.
