Patient Simulation TrainingReal-time voice · STT → LLM → TTSmedical.carecollaborative.cloud

Patient Simulation Training Architecture

An AI patient agent joins a live WebRTC room and role-plays a clinical scenario. Students conduct the encounter by voice while the platform transcribes, generates the patient's responses, and speaks them back in real time.

At a glance

What this use case is

Patient Simulation Training puts a medical student in a live voice conversation with an AI-driven patient persona. The student selects a scenario — a child with diabetes, an elderly patient with hypertension — and an AI agent joins the WebRTC room with scenario-specific persona instructions. The full STT → LLM → TTS pipeline runs in real time so the encounter feels like a real consultation.

Interaction Model
Real-time voice simulation
Deployed At
medical.carecollaborative.cloud
Data flow

Interaction pipeline

The ordered path a session takes through the platform for this use case.

1
Student voice → STT provider (real-time transcription)
2
STT output → AI Governance proxy → LLM (patient persona response)
3
LLM response → TTS provider (voice synthesis)
4
TTS audio → WebRTC room (returned to student)
Why it matters

Architecture highlights

Scenario-driven AI patient personas
Provider-agnostic STT / LLM / TTS pipeline
Template-driven communication assessment
Per-org Temporal namespace isolation
Service map

Platform capabilities

Internal Services
API Server (REST / DRPC / WS)
Temporal (durable orchestration)
Embedded WebRTC media server
AI Governance (MITM proxy)
Grading Engine (template-driven)
External Services (Provider-Agnostic)
Cloudflare (CDN / LB / Pages)
Identity Providers (SSO)
STT Providers (pluggable)
LLM Providers (pluggable)
TTS Providers (pluggable)
Go deeper

Explore the architecture

User Lifecycle — Patient Simulation Training

Every phase, route, and protocol for this use case, end to end.