VisionAery Platform · Aery Device Manager
On-Prem Control PlaneOperate edge AI across the whole fleet — from one console you own
The fleet management plane behind every VisionAery deployment. Mass model and configuration push, staged rollouts, per-site tuning, OTA updates, and continuous device health — running on hardware you own, behind your firewall. No SaaS in the alarming path.
The Problem
Edge AI is only useful if you can run it across a fleet
A single edge box on a single pad is a science project. A hundred boxes across forty sites in four regions is an operations problem — and it's the one most industrial AI deployments quietly fail at. Models drift, configs fork, a scene that worked last quarter doesn't this quarter, and nobody has the audit trail to explain why. Truck rolls multiply, and the alarming path becomes the part everyone is afraid to touch.
The Aery Device Manager is the control plane that makes large edge-AI fleets actually operable — staged model and config push, per-site tuning, OTA updates, and continuous health telemetry. It runs on hardware you own, behind your firewall: no VisionAery SaaS in the alarming path, no operator video leaving your network, no vendor account holding the keys to your fleet.
Devices
248
online
Sites
63
regions: 4
Uptime
99.97%
30d
Live Fleet
How Rollouts Work
A fleet-wide change, without the fleet-wide risk
Every model build, tuning change, and config update moves through the same staged pipeline — a small first batch, health-gated batches, and a one-click path back. Nothing hits the whole fleet at once, and nothing advances until the last batch proved healthy.
Stage
Pick a versioned model or config and a target — one site, a region, a tag-based group, or the whole fleet.
Test batch
The update goes to a small first batch of devices — never the entire fleet at once.
Health gate
Devices must clear post-deployment health checks before the next batch is allowed to advance.
Advance
Batches roll forward automatically as each one reports healthy, with a full audit trail.
Rollback
A regression or failed check reverts in seconds — one click, no truck roll.
Capabilities
Six capabilities that define the plane
Everything an operations team needs to deploy, tune, update, and audit an edge-AI fleet across many sites — from one console, on hardware you own.
01Mass model & config push
Push a new model build, tuning change, or config update to one site, one region, or the entire fleet from a single pane. Every push is rolled out to a small first batch of devices, versioned, and reversible — a regression rolls back in seconds, not a truck roll. Update windows, blackout dates, and per-region staging are first-class controls.
02Per-site, per-camera analytic tuning
Every scene is different — a stack flare isn't a refinery flare, a snowy tank at night isn't the same tank at noon. Operators tune sensitivity, dwell time, ROI masks, and time-of-day and weather-conditioned thresholds per camera from the console, no tech on the pad. Tuning ships through the same staged pipeline as model updates.
03Fully on-prem control plane
The management plane runs on hardware you own, behind your firewall — no VisionAery SaaS in the alarming path and no operator video leaving your network. It integrates with your existing identity provider (Active Directory / LDAP / SAML) and security stack as a tenant inside your operations network, not a service on someone else's infrastructure.
04Device health & OTA updates
Every box reports CPU, GPU, thermal, storage, network, and model state in real time. Firmware and software ship over-the-air through health-check gates: a device must report healthy before the next batch advances, and a failed post-update check rolls back automatically. Out-of-policy devices flag for review before they become an alarm-path failure.
05Version control for the edge fleet
Model versions, analytic configs, and per-site tuning are version-controlled artifacts. Every change carries an author, timestamp, reason, and a diff against the previous state — a site that worked last quarter and doesn't this quarter has an auditable history, not a war story. That versioning is what makes multi-region fleets safe to operate.
06Security posture by default
Edge devices ship with disk encryption, signed firmware, certificate-based mutual TLS to the control plane, and a default-deny network posture. Inbound from the WAN is closed; outbound is restricted to the on-prem control plane and approved destinations — aligning cleanly with IEC 62443 zone-and-conduit segmentation, enforced at the firewall.
Frequently asked questions about the Aery Device Manager
Related — what runs on top of the fleet
VisionAery Analytic · Flagship
Liquid Leak Detection
The flagship VisionAery analytic deployed and operated through the fleet management plane — with a working live demo.
VisionAery Service
Site Survey & System Design
The engineering work that designs the deployment the fleet management plane goes on to operate at scale.
VisionAery Analytic
Fire & Smoke Detection
Another analytic deployed and operated through the fleet management plane — early fire and smoke detection on existing cameras.
Make your edge AI fleet actually operable
Send us a fleet sketch — site count, devices per site, region layout, and the analytic mix. We'll come back with a deployment architecture and a real number, usually within one business day.