HomeSafetyLicensingTechnologyPartnershipCIRIS.ai

The Technology

Domain-routed medical AI built on the proven CIRIS framework

One Line of Custom Code

CIRISMedical's entire medical implementation consists of a single line that clears medical prohibition when all safety requirements are met. Everything else is pure configuration using CIRISAgent's existing capabilities.

WiseBus.PROHIBITED_CAPABILITIES = set() # The ONE line

Wisdom-Based Deferral (WBD)

Automatic Escalation to Humans

CIRISMedical automatically defers uncertain or high-risk decisions to medical professionals. The AI knows what it doesn't know and actively seeks human guidance.

Immediate Deferral Triggers

  • • ESI-1 or ESI-2 triage levels
  • • Confidence below 80% threshold
  • • Conflicting medical indicators
  • • Pediatric or pregnancy cases
  • • Medication interactions detected

Deferral Process

  • • AI flags case for review
  • • Notifies on-call physician
  • • Provides full context and reasoning
  • • Waits for human decision
  • • Logs entire decision chain

Mixed Modality Support 🚧

Status: In Development - Architecture supports modality extensions through wisdom providers. Text processing implemented, other modalities in progress.

Planned support for multiple data types to create complete health picture:

Text (✅) & Voice (🚧)

  • ✅ Text symptom descriptions
  • 🚧 Voice transcription
  • 🚧 Multi-language processing
  • ✅ Medical history intake

Sensor Data 🚧

  • 🚧 Vital signs monitoring
  • 🚧 IoT medical devices
  • 🚧 Wearable health trackers
  • 🚧 Environmental conditions

Visual Input 🚧

  • 🚧 Wound assessment photos
  • 🚧 Rash identification
  • 🚧 Swelling evaluation
  • 🚧 Simple diagnostic imaging

Offline-First Architecture 🚧

Status: In Development - Core architecture supports 4GB RAM constraint, offline capabilities being implemented for medical deployment.

Planned Edge Features

  • • Local model execution on 4GB RAM
  • • Sync when connection available
  • • Battery-aware processing
  • • Works on basic hardware
  • • Solar-powered device compatible

Planned Caching

  • • Medical protocols cached locally
  • • Common medication database
  • • Triage decision trees
  • • Emergency procedures
  • • Regional disease patterns

Privacy & Data Protection

Data Sanitization

Log sanitization and adaptive filtering protect sensitive information:

✅ Implemented

  • • Log sanitization for emails/usernames
  • • Control character removal
  • • Adaptive content filtering
  • • Secrets management system

🚧 In Development

  • • Full PII auto-redaction
  • • Medical record anonymization
  • • Geographic data masking
  • • Temporal shifting for dates

Comprehensive Audit Trail ✅

Audit Retention

Every decision is logged with configurable retention:

  • • 90-day default retention (pilot)
  • • Complete decision chain
  • • AI reasoning process
  • • Human interventions
  • • Deferral decisions
  • • Version history
  • • Model used for each decision
  • • 7-year retention planned for FDA compliance

Immutable Records ✅

Cryptographically signed audit trail ensures accountability:

  • • Ed25519 digital signatures
  • • Hash chain verification
  • • Tamper-evident storage
  • • Graph-based persistence
  • • Export capabilities (JSONL, CSV, SQLite)
  • • Third-party verification support
  • • Compliance reporting

DSAR & Data Rights ✅

Data Subject Access Requests - Implemented

GDPR-compliant data rights management system with API endpoints:

Available Rights ✅

  • • Access personal data
  • • Request corrections
  • • Export data
  • • Delete requests
  • • 14-day response time (3 for urgent)
  • • Ticket tracking system

Processing Features

  • • API endpoint: /v1/dsr
  • • Unique ticket IDs
  • • Status tracking
  • • Email notifications
  • • Audit logging
  • • In-memory storage (pilot)

Transparency & Accountability ✅

Public transparency feed implemented at /v1/transparency

✅ Public Feed (No Auth)

Live anonymized statistics:

  • • Total interactions processed
  • • Action breakdown (SPEAK, DEFER, REJECT)
  • • Deferral statistics
  • • Safety metrics
  • • Response times
  • • DSAR request counts

✅ System Metrics

Real-time health monitoring:

  • • Uptime percentage
  • • Average response time (ms)
  • • Active agents count
  • • Rate limit triggers
  • • Emergency shutdown attempts
  • • Harmful requests blocked

Emergency Response Protocol 🚧

Status: In Development - ESI triage framework planned, emergency protocols being designed with medical advisory board.

Planned Emergency Features

ESI Triage (Planned)

  • • 5-level severity assessment
  • • Automatic escalation protocols
  • • Emergency personnel alerts
  • • Transport coordination
  • • Location services integration

First Aid Guidance (Planned)

  • • Step-by-step instructions
  • • CPR guidance
  • • Bleeding control
  • • Stabilization protocols
  • • Communication maintenance

Cultural & Language Support 🚧

Status: In Development - Working with local communities and medical anthropologists to ensure culturally appropriate care.

Planned features for culturally sensitive healthcare:

Language Support 🚧

  • • Multi-language interface
  • • Local dialect adaptation
  • • Medical term translation
  • • Voice input support

Cultural Integration 🚧

  • • Traditional medicine respect
  • • Local healer collaboration
  • • Cultural practice awareness
  • • Community engagement

Religious Sensitivity 🚧

  • • Dietary considerations
  • • Gender preferences
  • • Prayer timing awareness
  • • End-of-life customs

Built for Scale, Designed for Safety

CIRISMedical combines proven AI technology with rigorous safety protocols, creating a medical AI system that's both powerful and responsible.

Explore Safety ArchitecturePartner With Us