Your edge.
Duplex on your own devices, on CPU.
Duplex runs where the work happens: on the machines, vehicles, sites and devices you own, on ordinary CPUs with no GPU. It works beside the AI already on the device, keeps learning and deciding with no connection, and nothing has to leave.
- Operated by you
- Runs on CPU, with no GPU
- Beside your edge AI
- Works offline, even air-gapped
- Enterprise
Beside your edge AI.
The model on your device handles what is new. Duplex answers what it has already learned, on the device’s own CPU, and learns from each result your team confirms, right there.
A crack like many before: Duplex rejects the part on the spot, on the camera’s own CPU.
On CPU, with no GPU.
Duplex needs no GPU of its own. It runs on the ordinary processors already in your devices, machines and on-site servers, and leaves any accelerator to your edge AI.
A reading arrives on the device, from a sensor on the line.
It learns on the device.
An edge model learns only when it is retrained somewhere else and shipped again. Duplex learns from each result your team confirms, on the device, as the work happens, so the next request already benefits.
A new kind of fault appears on the line for the first time.
With or without a connection.
Duplex learns, decides and acts on the device itself, so a dropped link, a remote site or an air gap does not stop it. It needs nothing outside to start, and its licence works offline.
In port, the vessel’s systems are online for the last time before sailing.
Nothing has to leave.
Data, what Duplex learns and its history can stay on the device or on your site. With a local model, nothing leaves at all: Duplex calls an outside model only if you set one up.
The readings and requests it handles stay on the device.
One device, or a whole fleet.
Each device signs in under its own name and works from a short list your team sets. Change a rule once, and every device works from the new list. What one device learns can reach the others when you allow it.
A new pump on site signs in under its own name, and says what it can do.
Where the work happens.
Wherever there is a CPU and work to learn from, Duplex can run beside the AI that is already there.
Factories and plants
Inspection, packing and process lines that decide on the spot and learn every shift.
Vehicles and fleets
Vans, trains and machines that keep deciding through tunnels, dead spots and long routes.
Stores and branches
Checkouts, lockers and back offices that learn local exceptions without sending data away.
Energy and utilities
Turbines, substations and pumps that act within their limits and send the rest back.
Clinics and labs
Instruments and on-site systems where the data has to stay in the building.
Ships, mines and remote sites
Places with no connection for days or weeks, where the work still has to go on.
What it runs on.
Planned with you for each deployment, on hardware you already have.
An ordinary CPU
Duplex needs no GPU, on a device, a machine or an on-site server.
Your edge AI, or a local model
Duplex works beside the model on the device, or a local model on your site.
Offline licensing
Its licence works without a connection, so nothing outside has to be reachable.
Signed releases
Every release is signed, so your team can check it before it goes onto a device.
Your own identity provider
People sign in through your own identity provider, and every device under its own name.
Local dependencies
Everything Duplex needs to start is on the device or your site.
Plan your deployment with us.
Duplex runs in a cloud account you own, or on your own devices and machines at the edge, on CPU. We plan each deployment with you. Contact us at hello@vareon.com.