← All Case Studies
Medical Devices Healthcare

Offline AI voice interface
for a medical device

Embedded AI Interface Turnkey Product Development
Background

Operating the device previously required a dedicated "circulating nurse" from the OR team – due to the skilled labor shortage, this role is increasingly hard to fill. At the same time, the surgeon cannot manually operate the device for sterility reasons. A cloud-based voice control was ruled out: the OR has strict requirements for data privacy, data security, latency, and failsafe operation.

What we built

A fully offline voice interface as a standalone hardware module, enabling voice control of all device functions.

Electronics development: microphone array, baseboard for System-on-Module, speaker module with Class-D amplifier
Electromechanical / thermal integration into an existing medical device – in close collaboration with the customer
Optimized voice capture for acoustically demanding environments – via classical signal processing and AI
Embedded AI pipeline: models for speech recognition, language understanding, and speech synthesis run fully locally
Optimization / quantization of AI models for inference on resource-constrained hardware
No cloud dependency, no latency from external APIs, no data privacy concerns
Integration with the device's native control system
Robust, protected, and safe design of the overall system for continuous clinical use
Result

The system has been running continuously in a clinical trial phase in the OR of a university hospital for over a year. Given positive feedback from the OR team, the manufacturer plans to offer it as an add-on module in the next device generation, which is currently moving toward series production.

Customer info
Location:
Germany
Industry:
Medical Devices · Healthcare

The project is in active development and clinical testing. At the customer's request, customer and product details remain confidential until launch.

Tech stack
Custom Electronics Microphone Array Beamforming Offline ASR Offline NLU Offline TTS ARM64 SoC System-on-Module Docker Kubernetes TensorFlow Lite Custom Interface Acoustic Measurement
Discuss a similar project →