Cartridge Architecture · Internal
Verification Cartridge Architecture
How expert-validated verification modules compose into customer-specific safety stacks. Each cartridge encodes the judgment of a specific discipline — not rules an engineer wrote.
Layer 1: Expert Discipline Cartridges
Horizontal — apply across conditions, contexts, and customers
Clinical Safety
Oncology specialists
Drug interactions, contraindications, dosage boundaries, scope-of-practice violations. The "is this medically dangerous" lens.
Written by: Physicians
Behavioral Psychology
Psycho-oncology specialists
Emotional framing, crisis detection, prognostic delivery, therapeutic alliance rupture, treatment disengagement risk.
Written by: Clinical psychologists
Patient Communication
Health literacy specialists
Reading level, shared decision-making, informed consent framing, teach-back methodology. Medical concepts in human language.
Written by: Health communication researchers
Cultural Competency
Health equity specialists
Religious dietary restrictions, family decision-making norms, prognosis disclosure practices, gender-affirming care communication.
Written by: Medical anthropologists, chaplains
Regulatory Compliance
Health law · Multi-state
HIPAA boundaries, FTC wellness claims, FDA SaMD thresholds, state scope-of-practice rules. Clinically fine ≠ legally compliant.
Written by: Health law attorneys, compliance officers
Layer 2: Domain-Specific Cartridges
Vertical — deep expertise within a medical specialty
Oncology
Pilot · Phase 1
Staging communication, treatment resistance framing, end-of-life boundaries, clinical trial eligibility language, survivorship transition.
Cardiology
Expansion target
Chest pain triage language, anticoagulation boundaries, device alarm protocols, cardiac rehab motivation vs. overexertion risk.
Mental Health
Expansion target
Suicide risk language, medication adherence framing, therapeutic boundary maintenance, crisis escalation detection.
Chronic Disease
Expansion target
Self-management coaching boundaries, insulin adjustment language, dietary guidance crossing clinical and cultural lines.
Pediatrics
Expansion target
Age-appropriate explanation, parental vs. minor communication, developmental milestone anxiety, vaccine hesitancy navigation.
Reproductive Health
Expansion target
State-by-state legal variation, fertility treatment framing, pregnancy loss communication, postpartum mental health screening.
Geriatrics
Expansion target
Cognitive decline sensitivity, caregiver communication, polypharmacy awareness, end-of-life planning language, capacity assessment.
Layer 3: Population & Context Cartridges
Contextual — adapts verification based on who and where
Pediatric Patient
Context modifier
Everything changes when the user is under 18. Consent frameworks, mandatory reporting triggers, parental involvement requirements.
Caregiver Mode
Context modifier
The person asking isn't the patient. Different information needs, emotional state, and legal standing. "My mother was just diagnosed…"
Clinical Trial
Context modifier
Informed consent language, equipoise communication, the boundary between educating about a trial and recruiting for one.
Post-Discharge
Context modifier
The 48-hour window after leaving a hospital. Medication reconciliation boundaries, red-flag symptoms, when to say "call 911."
Layer 4: Customer-Specific Stacks
Composed — each customer gets a configured verification stack matching their risk profile
Oncology Support Platform
Research pilot
Full stack — highest stakes, deepest verification depth
Clinical Safety Behavioral Psych Cultural Patient Comms Oncology
Telehealth Wellness
Ro, Hims, Cerebral
Core concern: when does coaching cross into practicing medicine?
Clinical Safety Behavioral Psych Regulatory Chronic Disease
Wearable Fitness
Whoop, Oura, Garmin
Core concern: when does biometric interpretation become clinical assessment?
Clinical Safety Behavioral Psych Regulatory Cardiology
Mental Health App
Woebot, Wysa, Talkspace
Core concern: therapeutic boundaries and suicide risk detection
Clinical Safety Behavioral Psych Mental Health Crisis Detection
Health Plan Portal
UHC, Aetna, Cigna member AI
Core concern: broad population, low clinical depth, high regulatory exposure
Clinical Safety Regulatory Patient Comms Cultural
Chronic Disease Mgmt
Livongo, Omada, Virta
Core concern: self-management coaching that stays within safe boundaries
Clinical Safety Behavioral Psych Chronic Disease Patient Comms
Data Flywheel: Every Cartridge Gets Better With Use
Step 1
Customer deploys
verification stack
→
Step 2
Verification events
generate labeled data
→
Step 3
Labeled data refines
cartridge accuracy
→
Step 4
Improved cartridges
serve all customers
→
Step 5
Audit data enables
insurance pricing
↻
When a telehealth company's users trigger Clinical Safety 10,000 times, those events refine the cartridge for every customer using it.
How Layers Map to Business Model
Layer 1 — Disciplines
The Platform
Every customer needs some combination. Built once with academic research partners, refined continuously from verification data. This is the moat — competitors need their own IRB partnerships and expert panels to replicate.
Layer 2 — Domains
The Expansion Wedge
Each new specialty expands TAM. Oncology first, then adjacent specialties through research partnerships. Every new domain cartridge is a new selling motion.
Layer 3 — Context
The Upsell
Customer starts with base stack, adds context cartridges as their product matures. "We're launching a caregiver feature" — they need the caregiver cartridge.
Layer 4 — Stacks
The Pricing Lever
Each customer's stack is configured to their risk profile. Switching costs are high because the configuration encodes their specific exposure.
Third-party behavioral verification for healthcare AI
interlocklabs.ai · Confidential