Architecture
How it's built. Today and next.
We separate live capabilities from the pilot-phase roadmap so clinics know exactly what they get today and what lands during the pilot.
Live today
LiveWeb application
React 19 + TanStack Start on Vite. SSR-capable, deployed on edge.
Data plane
Managed Postgres with row-level security scoped per clinic.
Storage
Private object storage. All patient imagery served through short-lived signed URLs.
Auth
Managed authentication with email/password and Google OAuth.
Analysis engine
Vision inference through the Lovable AI Gateway (Gemini).
Audit
Every mutating action written to an immutable audit log per clinic.
Planned — pilot phase
RoadmapAWS S3 object store
Migrate image storage to per-region S3 with lifecycle rules.
Amazon SageMaker inference
Fine-tuned dermatology classifier hosted on managed inference.
AES-256 at rest / TLS 1.3 in transit
Formalise encryption certifications during pilot deployment.
HIPAA-adjacent controls
BAA-ready deployment tier during pilot with a partnered dermatology group.
DICOM ingest
PACS bridging for clinics with existing imaging infrastructure.
SOC 2 Type I
Audit engagement kicked off after the first paying pilot.
Regulatory posture. Aimazing is not currently classified as a medical device. AI outputs support clinical judgement and are not diagnoses. During the pilot programme, we work with participating clinics on compliance controls appropriate to their jurisdiction.