Case study · AI in production
AI Clinical Genetics Platform with Genetic Test Ordering
A dedicated WeyBee team modernised an AI clinical genetics platform used by 10,000+ hospitals and clinics: 18 microservices, a cross-platform mobile app and end-to-end genetic test ordering, with security and patient-data privacy built into every release.
AIThe client
The client provides AI decision-support tools that help clinicians evaluate patients for developmental and genetic disorders. Clinicians upload patient photos and clinical findings, and the platform's deep-learning engine suggests possible syndromes to consider. The platform supports clinical judgement and does not replace it. It is used by more than 10,000 hospitals, clinics and organisations worldwide.
The challenge
- Break a legacy monolith into maintainable services without disrupting clinicians who use the platform every day.
- Move thousands of existing users to a new identity provider without forcing a password reset.
- Add end-to-end genetic test ordering: patient data, documents, consents, a generated letter of medical necessity and submission to labs.
- Run notification campaigns, batch case processing and surveys at scale.
- Protect patient photos, clinical findings and test orders, which are personal and protected health information (PII and PHI), in line with HIPAA safeguards and GDPR requirements.
What we did
We provide a dedicated team to the client. The team split the legacy monolith into domain microservices using Domain-Driven Design, placed Backend-for-Frontend layers in front of them, and built three front-end applications, including a cross-platform mobile app. Everything runs on AWS, with Kubernetes and serverless components managed as infrastructure-as-code and GitOps pipelines. Security and data privacy run through all of it: we remediate CVEs and vulnerable dependencies across the services, handle PII and PHI with care, and build to the client's HIPAA safeguards and GDPR requirements.
What it does
AI facial analysis workflow
Clinicians create cases and upload photos; faces are detected and compared automatically, and the AI engine returns syndrome suggestions.
Clinical case management
Case lists and details, phenotype findings, chart notes, reported diagnoses and review status.
Genetic test ordering
Patient information, documents and images, e-consents, a generated letter of medical necessity (PDF) and submission to labs.
Secure identity
JWT login, TOTP multi-factor authentication, and a migration to a managed identity service with no forced password resets.
Serverless engagement platform
Batch case processing, segmented email and SMS campaigns, and forms and surveys.
Medical knowledge services
Syndrome library, questionnaires, consent management and research data APIs.
Security and data privacy
CVE and dependency remediation
We track security advisories (CVEs) for the libraries, frameworks and container images behind 18 microservices and three apps, and patch or upgrade vulnerable dependencies as part of regular delivery.
Careful handling of PII and PHI
Patient photos, clinical findings, consents and test orders are treated as protected health information, available only to authenticated, authorised users through the platform's secured services.
HIPAA safeguards
Our engineering supports the client's HIPAA safeguards: multi-factor authentication, a managed identity service, controlled access and a recorded trail of each step in the clinical workflow.
GDPR requirements
Electronic consents, consent management and research data APIs are built to the client's GDPR requirements for consent and access to personal data.
The client owns its compliance programme; our engineers build and maintain the software to its security and privacy requirements.
How it is built
- 18 backend microservices: 5 Backend-for-Frontend layers and 13 core domain APIs, behind 4 API gateways.
- 500+ REST endpoints across Laravel, FastAPI and Node.js services, with standard error handling.
- 22 AWS Lambda functions orchestrated with Step Functions, SQS, SNS and EventBridge.
- Kubernetes on EKS with Helm, Argo CD and Istio; infrastructure in Terraform and AWS CDK.
- MySQL, DocumentDB, Redis and OpenSearch; RabbitMQ and Celery for background work; Sentry monitoring.
- Turborepo monorepo for web and React Native mobile apps.
- Security in every layerTOTP multi-factor authentication, a managed identity service (Cognito), API gateways in front of every service, and continuous CVE and dependency remediation.
Results
- More than 550,000 patient cases analysed on the platform.
- More than 12,000 new cases a month, from 10,000+ hospitals, clinics and organisations.
- Legacy monolith replaced by domain services the client can extend independently.
- Users moved to the new identity service without a forced password reset.
- Security and privacy built into every release: vulnerable dependencies remediated, and PII and PHI handled to the client's HIPAA and GDPR requirements.
Engineering around a deep-learning
engine used in clinics
The client's deep-learning engine suggests possible syndromes from patient photos. Our team engineered the workflow that makes the AI usable in clinical practice: case intake, face detection and comparison, results, and the path to genetic testing.
Clinicians create a case and upload photos; faces are detected and compared automatically and the engine's suggestions return to the case.
Suggestions flow straight into genetic test ordering, with consents, documents and a generated letter of medical necessity.
Microservices, serverless batch processing and secure identity carry 12,000+ new cases a month for 10,000+ hospitals and clinics, with patient data protected throughout.
The workflow keeps the clinician in charge: AI output supports judgement and every step is recorded.