Workflow intelligence for clinical operations.
Six pre-built verticals. YAML workflow templates. Deployed entirely inside your cloud. This is how SmartFlowCraft Health works.
Templates are data, not code.
Each workflow is a YAML file describing a DAG of nodes. Adding a new vertical means writing a new template, not shipping Python.
Templates are version-controlled, readable by non-engineers, and auditable — a requirement for any regulated workflow.
name: retainer_order
version: 1
vertical: orthodontics
input_schema:
type: object
additionalProperties: false
required: [order_id, case_type, tenant_id]
properties:
order_id: { type: string }
case_type: { enum: [standard, complex, rush] }
tenant_id: { type: string }
nodes:
- id: validate
type: rule
rules:
- if: "case_type == 'rush' && !tenant.allow_rush"
then: { status: HOLD, reason: rush_not_enabled }
- id: assign_lab
type: rule
strategy: round_robin
source: tenant.labs
- id: emit_decision
type: http
target: "{{tenant.webhook_url}}/workflow-decision"
edges:
- { from: validate, to: assign_lab, when: "status != HOLD" }
- { from: assign_lab, to: emit_decision }The compliance boundary is your cloud, not ours.
Your cloud (client data plane)
Patient records
STL files, scans, PDFs
Order history
PHI of any kind
LLM inference (your Bedrock / Azure OpenAI / Vertex AI)
Application runtime
Databases
SmartFlowCraft (control plane only)
Template library
Workflow configuration
Run counts (metadata)
Template version tracking
Billing and metering
Deploy orchestration
PHI never reaches this side.
One image. Three clouds.
AWS
IaC
Terraform
Services
ECS Fargate, RDS Postgres, Elasticache, S3, Bedrock
LLM (BAA-covered)
Claude via Bedrock
Azure
IaC
Terraform or Bicep
Services
Container Apps, Azure DB for PostgreSQL, Cache for Redis, Blob Storage
LLM (BAA-covered)
GPT-4 via Azure OpenAI
GCP
IaC
Terraform
Services
Cloud Run, Cloud SQL, Memorystore, Cloud Storage, Vertex AI
LLM (BAA-covered)
Gemini via Vertex AI