From Days and Weeks to Minutes: MMS Transforms Information Retrieval With APRÓ's Secure AI Solution

See how MMS deployed APRÓ's secure AI solution, cutting information retrieval from hours to seconds using AWS Bedrock and RAG.

Articles

Better access to school information with secure and practical AI.

APRÓ partnered with the Ministry of Education and School Services (MMS) to implement a secure and practical AI solution built on AWS Bedrock and a customized LibreChat interface. This solution enables MMS staff to use advanced AI tools like Claude for daily tasks and to access information from internal databases in a secure and simple way.

The Challenge

MMS needed to implement an AI solution that guaranteed:

  • Compliance with GDPR and Icelandic data protection laws.

  • Secure integration with their sign-in system (SSO).

  • That no data would be used to train AI models. A solution that could scale with the institution's growing needs.

APRÓ's Solution

We developed a customized LibreChat interface, integrated with MMS's SSO, and powered by AWS Bedrock to provide secure access to foundation models.  APRÓ is using open source software that works like a single unified chat interface for many different AI models. Instead of needing to open ChatGPT, Claude, or other external AI assistants each time, MMS staff can use one interface for all AI-related tasks. For MMS, this meant we could create a custom solution that securely connects both external AI models (through AWS Bedrock) and MMS's internal data.

 The solution offers:

  • General AI tools for summaries, translations, and content creation.

  • Specialized "agents" that use Retrieval-Augmented Generation (RAG) to access MMS data.

  • Role-based access controls for internal and external users.

  • Full encryption and audit trails for all operations.

Technical Setup

  • Interface: Customized LibreChat (APRÓ OSS), hosted on AWS Amplify, protected with CloudFront and WAF.

  • Backend: API Gateway + Lambda for orchestration and security policies.

  • Data layer: Amazon S3 and Amazon OpenSearch Service for RAG.

  • AI layer: AWS Bedrock for secure access to foundation models without training on MMS data.

  • Security & compliance: AWS IAM Identity Center, AWS KMS, CloudWatch, and VPC endpoints.

Results & Outcomes

  • Time to find internal documents decreased from hours to seconds.

  • Improved productivity in content creation and planning.

  • 100% GDPR compliance with zero risk of data leakage.

Time savings and improved workflows at MMS

The AI solution that APRÓ delivered has transformed our workflow at MMS. Now we can use AI in a secure way, from planning to accessing information in our own databases." - Páll Ásgeir, IT Director, MMS

„Collaborating with MMS was really enjoyable, and it was great to see how AI can be put to use at a public institution in a secure and practical way. This project shows how important it is to implement such solutions with the right security measures in place." - Jón Levý Guðmundsson, Team Lead at APRÓ.


Contact us