Our customers share their stories of successful collaboration
Case
Delivering scalable GenAI capabilities to healthcare organizations with an enterprise-grade service platform
Cambrian designed, built and deployed a GenAI service platform allowing the individual tenants of this software company to consume Amazon Bedrock based features in a controlled manner.
The service platform offers an intelligent cost management system, multi-tenant support and enterprise grade security and compliance. A tiered budget system provides cost certainty for Adapcare’s end-clients and real-time cost tracking, while the new features have reduced their administrative overhead by 90%.
“They are great to work with, have the right expertise, and are available night and day when we need their support.”
Ramon Postulart – CTO
The customer
Adapcare builds software to support healthcare and social care providers. The company provides their electronic healthcare records platform, along with scheduling and shift management systems, to 50,000 health professionals.
The challenge
Healthcare systems face a structural shortage of skilled personnel. One way of relieving the associated pressures is to reduce the significant administrative workload on the healthcare professionals. Generative AI is a technology that can be used for this purpose but it needs to be deployed in a compliant, secure and scalable manner. At the same time, providing sufficient cost control mechanisms to healthcare providers is essential for a successful market adoption.
Adapcare was looking for a way to provide their healthcare customers with reliable access to their new large language models. These were designed to aid in mandatory patient documentation. Pilots had shown that the administrative burden of the healthcare professionals for these tasks could be reduced by as much as 90%.
In order to launch their new platform features successfully, it was key to maintain strict cost controls, ensure service reliability, and meet specific healthcare compliance requirements. However, the traditional synchronous AI implementations could not handle the scale and cost predictability required for enterprise healthcare deployments.
The solution
We designed and built a service component meeting these requirements.
This solution enabled each healthcare provider tenant to call the Amazon Bedrock API responsibly and it delivered the necessary functionalities to go-to-market successfully.
The key functionalities of this service component are:
- Intelligent Cost Management System－Each healthcare provider selects a monthly budget for their use of GenAI. The component tracks the usage costs per provider in real time with automatic currency conversion from USD to EUR, delivers proactive notifications at 90% budget utilisation, generates automatic budget renewal with zero service interruption and creates a comprehensive audit trail for all AI usage and costs.
- Multi-Tenant Architecture－The component has customer-specific model access controls via Application Inference Profiles, granular permissions management with API key authentication, isolated processing environments per individual customer and configurable model aliases for simplified client integration.
- Enterprise-Grade Reliability－An asynchronous processing architecture prevents timeouts, enables fault-tolerant job queuing with automatic retry mechanisms, inlcudes comprehensive error handling and logging, and allows for multi-environment deployment (dev, staging, production).
The technical characteristics of the implementation are:
- Infrastructure as Code－Consistent deployment is ensured with complete infrastructure automation, supporting environment-specific configurations and with automated deployment via CICD pipelines.
- Microservices architecture－Separate handlers for async processing, budget management, and notifications were built in with an event-driven design for communication.
- Security & Compliance－Characteristics include least privileges role-based access controls for all AWS services, data encryption in transit and at rest, API rate limiting, CORS configuration, API key management, and comprehensive CloudWatch logging for compliance tracking.
The results
Since go-live, the solution has comfortably met the 99.9% availability SLA, with sub-second API responses, zero budget overruns and a seamless promotion from acceptance to production.
The tiered budget system provides cost certainty for Adapcare’s end-clients, is reducing the administrative overhead by 90% and provides real-time cost tracking.
The serverless architecture automatically handles demand spikes by autoscaling. It also allows the flexible selection of the foundation model based on use case requirements. Due to the asynchronous processing, long-running AI tasks are handled without blocking the client applications.
Finally, the multi-region capability for some existing international healthcare clients was built in, which also supports the use of the AWS Sovereign Cloud in Europe.