An Australian mental-health practice had already integrated Claude into its clinical workflows but faced privacy-compliance hurdles under Australian data-sovereignty law. Infostatus was engaged to recreate the experience inside a secure, compliant, AWS-hosted application, removing the need for manual de-identification and enabling smoother clinical workflows.

The challenge

• Clinicians manually de-identified and re-identified sensitive notes to use AI
• Generating assessment letters took several manual steps
• Scanned documents could not be processed because of privacy restrictions
• There was no assurance that model processing stayed within Australia

What we did

We built a secure, Claude-style chat and project interface as a containerised application, deployed to Amazon ECS on Fargate behind an Application Load Balancer, with model inference through Amazon Bedrock using Claude in the Sydney region.

• Networking and delivery: Amazon Route 53 and an Application Load Balancer with HTTPS termination
• Compute: Amazon ECS on Fargate, with images stored in Amazon ECR
• AI: Amazon Bedrock with Claude, scoped by least-privilege IAM task roles
• Storage and audit: Amazon S3, AWS CloudTrail and Amazon CloudWatch, with Amazon GuardDuty for threat detection

Every resource was deployed to the Sydney region for full data sovereignty, with encrypted endpoints and audit-ready logging.

Outcomes

• Around 40% reduction in administrative overhead
• Manual de-identification eliminated
• AI analysis of scanned and handwritten documents enabled
• Full Australian data residency with no compliance ambiguity
• A platform ready to support additional clinical workflows