Work · Health · Voice AI · VOX/04 / 2024 → 2025
Sovereign voice agents for the care pathway
A platform that hands a practice’s repetitive tasks (booking, changing, cancelling appointments, answering questions) to conversational agents plugged into existing business software. Voice answers in under 1.2 seconds and data stays in HDS-certified infrastructure: no proprietary American model in the loop.
The context
Medical practices lose hours on the phone for simple gestures: booking, moving, cancelling appointments, answering recurring questions. Patients call when the lines are saturated; the front desk does what it can. And health data imposes a strict frame: HDS-certified hosting, and a sovereignty requirement on the models.
The work
Design and delivery of the complete platform, as technical founder: VOIP gateways, streaming speech recognition and synthesis, orchestration of conversational agents plugged into existing practice management software. End-to-end latency under 1.2 seconds, the threshold where a voice stays natural. Inference on GPUs provisioned on demand, open models only: no health data ever crosses a proprietary American model.
What remains
Agents in production that relieve teams of repetitive tasks without degrading how patients are received. And proof that demanding medical voice AI (real-time, sovereign, compliant) can be built without architectural compromise.
- End-to-end voice latency under 1.2 s
- Encrypted infrastructure, HDS-certified hosting
- Sovereign inference: on-demand GPUs, open models
- Native integration with practice management software