A prototype in development: helping elderly and frail people, driven by voice, with the AI running on board and the data staying in the house.
Or write to us: info@ggtechnologies.sm
Anyone caring for an elderly parent would like to know whether they have fallen, whether they have eaten, whether the house is too cold. The tools that make that possible — cameras, microphones, sensors — are also the ones nobody wants in the living room, because they send everything to somebody else's server.
We are building a prototype that tackles the contradiction on the technical side: the robot sees and hears, but the model runs on board and what it recognises stays in the house. It is DigiSense® applied to the most delicate case we know. Today it is a prototype in development, not a product you can buy.
A prototype in development. It is not a product for sale and it is not a medical device.
On board the robot, on the DigiSense® framework: audio and video never leave the house.
Reachy Mini, the open-source platform by Pollen Robotics (Hugging Face group).
We start from Reachy Mini, the open-source robotic platform by Pollen Robotics, rather than building the mechanics from scratch: the work goes into the capabilities.
A voice module and an internal LLM: you talk to the robot in plain words and the answer is formed on the machine, without asking the internet for anything.
Camera, temperature, humidity and microphone. The microphone is being trained to recognise sounds that may indicate a fall or a call for help.
A working prototype on Reachy Mini, with the voice module, the on-board LLM and the ambient sensors.
Recognition of sounds that may indicate a fall or a call for help, and refinement of the voice interaction.
A home-automation hub connecting the robot to the wearables already in the house, to read their vital-signs data.
Recognising a person from their electrocardiographic signature. That is biometric data: before the engineering, we have to settle how to handle it.
No. It is a prototype we are working on, not a catalogue product. If you would like to follow its development or host a field trial, write to us.
We do not say that, and the distinction matters. We are training the microphone to recognise sounds that may indicate a fall or a call for help: it is an aid, not a safety system to rely on. A device that promises to detect falls is a different thing, with different obligations.
In the house. The model runs on board the robot and the recognition happens there: no server receives the camera or microphone stream. That is why we chose this architecture instead of a simpler one in the cloud.
That a heart trace is distinctive enough to tell one person from another, so a wearable could tell the robot who is in front of it without using the camera. It is a future possibility, not a feature: it is biometric data, and how to handle it has to be settled before the engineering.