Edge Inference

Barvis Edge Node

A single box, sized for your room, sitting on your power and your network. Every frame is processed inside your building. What crosses the network is a stream of counts — timestamp, station, SKU, pegs, confidence — orders of magnitude smaller than video and carrying no imagery. There is no cloud video pipeline to turn on later.

GVEN-HW-0093In deployment

The problem

Most camera products in hospitality work the same way: the venue streams video out, someone else’s servers do the thinking, and the footage lives somewhere you cannot see. It is easier to build. It is cheaper to develop. And it puts your staff, your patrons and your liability in a data centre you do not control.

We did not want that answer, and at the scale this product has to reach we could not have afforded it anyway.

Overview

There are two reasons all real-time inference runs on a box inside your venue, and it is worth knowing both.

The first is arithmetic. A cloud GPU dedicated to a single venue costs more per month than the venue will pay for the entire product, and streaming 1080p out of a thousand bars means paying for a thousand uplinks that most venues do not have. Edge-first is not a preference at this scale. It is the only architecture that closes.

The second is that we are pointing a camera at people at work, in a country that has told us in statute how to behave about that. So the answer is structural rather than procedural, because a policy you can switch off is not an answer. Raw frames do not cross the boundary — there is no ingest endpoint for them to cross to. Egress is a JSON event with no imagery and no identifier of any person. And the event schema has no field for who, which means purpose limitation survives a change of management.

One box. Your power, your network, your building.

Who it is for

Nobody buys an edge node. They buy Barvis and the node comes with it. But it is the deciding factor in two specific conversations.

The privacy conversation

For an owner worried about their team, a hotel group with a compliance function, or anyone who has read a headline about workplace surveillance. Nothing leaves the building — which is a different claim from encryption in transit.

The IT conversation

For a hotel or chain where a security or IT team has to approve anything on the network. A device that egresses small JSON events and ingests nothing is a very easy approval compared with a video uplink.

What it does

Runs every real-time model locally

Detection, tracking, the pour trigger, volume estimation and event confirmation, all inside the venue.

Handles multiple cameras

Sized to the number of pour zones and context cameras in the room.

Buffers when the line drops

Events queue locally and sync on reconnect. Counting never stops.

Emits events only

Small, structured, no imagery.

Updates over the air

Model and software updates delivered remotely without a site visit.

Reports its own health

So a camera knocked out of alignment during a busy Saturday surfaces as a fault, not as a mysterious dip in the count.

Sizing

Compact

Small bar, single pour zone

Up to 3 cameras

Standard

Standard venue, multiple stations

4–10 cameras

Flagship

Multi-bar venue or large club

12+ cameras

Specifications

Placement

Inside the venue, on your power and network. Typically in the back office or a rack near the counter.

Network requirement

Modest. Events are kilobytes; there is no video uplink.

Data leaving the premises

{ timestamp, station, SKU, pegs, confidence }

Raw video egress

None. No ingest endpoint exists.

Local video retention

Configured and controlled by the venue, under the venue’s own signage and notice obligations.

Offline operation

Full. Events queue and sync on reconnect.

Updates

Over the air.

What it does
not do

It does not stream video anywhere.

It does not store footage on our infrastructure.

Any local recording belongs to the venue.

It does not run facial recognition.

That model does not exist in the stack.

It does not need a dedicated internet line or a static IP.

It does not stop working when your connection does.

Barvis counts every pour at the bar and reconciles it against the till

All products