Bilingual Clinical Voice Agent on a Hospital Wearable
A wristband-plus-voice-AI platform for hospitals, assisted living, and hospitality — combining Whisper fine-tuning, custom Hebrew NLP, and multi-agent clinical orchestration.
The Challenge
A disposable wristband had to integrate with a bilingual (Hebrew/English) AI backend for hospitals, assisted living, and hospitality — supporting 24/7 virtual nursing, patient identification, fall detection, indoor/outdoor location, and EMR/EHR integration at clinical latency and accuracy.
The Architecture
What we built.
Wearable Integration
Mobile + IoT platform with QR-coded wristbands, 3-month battery, IPx7 waterproofing, indoor/outdoor location awareness.
Voice Pipeline
Whisper + Whisper-Turbo fine-tuned on clinical terminology (HE/EN); OpenVoice for contextual clinical TTS; real-time WebRTC + SFU under one second of latency.
Agentic Clinical Core
LangChain + LangGraph multi-agent system (Conversational, Router, Monitor) with toolkit agents per clinical capability; Dicta + custom Hebrew fine-tuning; Bedrock Claude + Mistral on Kubernetes.
Business Value
What changed for the customer.
- Production bilingual clinical voice agent handling real patient interactions in Hebrew and English.
- Multi-hospital rollout with no custom deployment per site — one platform, many facilities.
- Automated human-clinician fallback preserves full context when agent confidence drops — continuous care, never dropped.
- Wearable + AI + EMR integration shipped as one stack instead of three vendors.
