Klinik — AI Patient Intake Voice Assistant Prototype
Voice-driven patient intake & triage agent prototype built for medical hackathons
Emergency room and clinic admissions face bottlenecks during initial intake because intake staff must manually transcribe spoken descriptions into rigid electronic forms while managing patient distress.
Developed a hackathon proof-of-concept voice assistant that captures spoken symptom descriptions in real time, parses observations into structured Pydantic schemas, and routes dialogue through a stateful LangGraph machine.
Demonstrated functional real-time speech capture and structured intake summaries during live hackathon judging, with deterministic safety guardrails preventing any medical diagnosis.
Key Architecture Components
- 1FastAPI WebSockets layer handling real-time audio streaming from the client
- 2LangGraph state machine coordinating intent classification, triage inquiry branching, and fallback safety guards
- 3Strict Pydantic response modeling ensuring output conforms to medical intake record formatting
- 4Safety guardrail node rejecting medical diagnoses and enforcing immediate referral to licensed professionals
Engineering Decisions & Business Rationale
LangGraph over linear prompt chains
Intake conversations are non-linear; patients re-clarify symptoms or backtrack. Graph-based state transitions allowed conditional routing without brittle prompt chains.
FastAPI with async WebSockets
Minimizing audio latency is paramount for conversational voice. Async event-driven handling delivered smooth turn-taking in live demonstration testing.
Challenges & Trade-offs
- •Prioritized transcription fidelity and low latency over running edge models locally, using hosted speech-to-text APIs with strict data sanitization filters.
- •Framed explicitly as an administrative intake assistant prototype rather than a diagnostic tool to avoid clinical liability and hallucination risks.
What I Learned
- ✓Stateful agent loops need deterministic interrupt handlers when users correct previously stated information.
- ✓Latency budgets must be budgeted strictly across speech-to-text, LLM inference, and text-to-speech synthesis.
Interested in a similar build for your operations?
Book a free workflow review or message Charles directly on WhatsApp.