Hardware · Edge AI
Every building system. One bridge. Your own AI.
Kaephor connects any sensor, meter or controller — KNX, Modbus, BACnet, DALI, M‑Bus, LonWorks and every major wireless standard — into a single rack‑mount gateway that runs AI on‑premises. No cloud dependency. No per‑point licence. The building’s data never leaves the building.
The problem
A building speaks twenty languages. Your AI shouldn’t need twenty translators.
Every floor is a patchwork of incompatible buses and proprietary controllers. The incumbents tax you by the point and pull your data to the cloud. Kaephor collapses all of it into one on‑prem bridge.
Islands of protocol
KNX for lighting, BACnet for HVAC, Modbus for meters, DALI for luminaires, LonWorks legacy — none of them talk to each other, and none of them talk to AI.
Cloud is a liability
Streaming a building’s live operational data to a foreign cloud is a security, latency and compliance problem — especially across the GCC. Inference belongs on‑site.
Metered to the point
Incumbents licence both points and devices, then bill a mandatory annual maintenance fee. Your cost grows every time you add a sensor.
The device
It bridges networks, not sensors.
A building’s sensors already sit on multi‑drop buses. Kaephor terminates the buses — one port fronts hundreds of points — and normalises everything into one clean model for the on‑board AI.
- 01ComputeDual-brain: a Linux applications processor paired with a real-time fieldbus MCU.
- 02Local AI uplink2.5GbE / USB4 to an AMD Ryzen AI Max inference node.
- 03SecuritySecure boot · TPM 2.0 · hardware secure element.
- 04UpdatesSigned OTA with A/B rollback across the fleet.
- 05ArchitectureHybrid-modular base board with pluggable protocol cards.
- 06Capacity~5,000 points comfortable · 10,000 ceiling · 250–500 devices.
Every standard in the building
Two form factors
One platform, sized to the site.
Kaephor Bridge · 1U
The standalone protocol bridge.
Deploy one per floor, riser or plant room; federate them all to a central AI over the building LAN.
- Form
- 1U · 19” rack
- Points
- Up to 10,000
- AI
- External node · 2.5GbE
Kaephor Appliance · 2U
Bridge and AI in one enclosure.
An AMD Ryzen AI Max inference engine runs models locally — the turnkey box for a single building.
- Form
- 2U · integrated AI
- Compute
- Ryzen AI Max · 128GB
- Data
- 100% on-premises
How it scales
From a corner office to an airport.
You never wire sensors to one box. You place a bridge per zone and federate to a central AI. Sensors-per-building scale linearly — one large building is a fleet of bridges.
| Building | ≈ Points | Bridges | AI nodes |
|---|---|---|---|
| Small office | 150–600 | 1 | 1 |
| Hotel · 300–500 rooms | 4k–12k | 2–3 | 1 |
| Class-A office tower | 5k–25k | 2–5 | 1–2 |
| Hospital | 15k–60k | 3–12 | 2 |
| Airport terminal | 30k–150k+ | 6–30+ | 2–3 |
| University campus | 50k–500k+ | 10–100+ | 2–3+ |
The cost case
On-prem AI included — at supervisory-stack economics.
5-year platform cost for a representative 10,000-point building, in SAR — supervisory layer plus AI analytics. Kaephor lands ~40% below the AI-capable Niagara + SkySpark stack — and against a supervisory-only install with no AI at all, the on‑prem AI layer is already in the rack for roughly a third more.
- No per-point licence.
- No mandatory annual maintenance. Incumbents run ~SAR 26,000/yr and it grows; Kaephor Care is a flat, optional subscription.
- On-prem AI, included.
- Fewer boxes.
The advantage compounds
The bigger the building, the wider the gap.
Their per-point tax compounds; our model is flat. Kaephor’s cost advantage over the AI-capable Niagara + SkySpark stack grows with every point on the network.
Indicative figures. 5-year comparisons are for a representative 10,000-point building and cover the supervisory + analytics platform layer only (field devices, cabling and integration labour excluded). Incumbent licensing and maintenance are quote-gated and estimated from public reseller pricing; Kaephor figures are engineering estimates pending firm supplier quotes. The interactive sizing calculator is an indicative model — AI energy savings assume SAR 3.5–7/ft²·yr energy by building type and up to ~10% AI-driven optimisation scaled by instrumentation density; Wi‑Fi AI sensing is a preview capability. Device images are design-concept renders. Prices in SAR.
Size your building
Your building. Your rack. Your payback.
Pick a building type and size, choose how granularly you sense it, and watch the fleet, the price, and the AI saving resolve in real time.
Commercial model
Flat, unmetered. No per-point tax.
Hardware is bought once and the full point capacity ships unlocked. Support and on‑prem AI updates are one flat subscription — never metered by point or device. Pricing is quoted through Kaephor’s delivery partners, sized to your building.
Kaephor Bridge
Universal protocol bridge — up to 10,000 points across every wired & wireless standard. Pairs with an AI Node.
- 20+ protocols
- 2.5GbE uplink
- Secure boot · TPM 2.0
Kaephor Appliance
Bridge + on-prem AI in one enclosure. AMD Ryzen AI Max inference built in — the turnkey box for a single building.
- Integrated AI · 128GB
- All Bridge features
- 100% on-prem
Kaephor AI Node
Standalone AMD Ryzen AI Max inference node + Kaephor platform. One node federates many bridges across a building.
- Ryzen AI Max · 128GB
- Local analytics & LLM
- Serves 10+ bridges
Kaephor Care
Support, signed OTA, and AI-model updates. Flat — no per-point licence, no mandatory SMA.
- Fleet OTA
- AI model updates
- SLA support
Hardware and Kaephor Care are quoted through Kaephor’s delivery partners — with integrator margin built in and channel tiers at 25 / 100 / 500 units. Protocol expansion cards priced per card. Ask your Kaephor partner for a sized quotation, or book a briefing below.
Why Kaephor
Built to own the building’s data.
- 01Everything, in one box20+ wired and wireless standards bridged into a single normalised model — the breadth no off-the-shelf gateway offers.
- 02AI on-premisesInference runs locally on AMD Ryzen AI Max. Low latency, full data sovereignty — a decisive advantage across the GCC.
- 03Flat economicsNo per-point licence, no per-device tax, no maintenance ransom. Your cost stops scaling with your sensor count.
- 04Secure & serviceableHardware secure boot, TPM 2.0 and signed OTA keep a distributed fleet trustworthy and up to date for a decade.
Bring the building under your own AI
Let’s put a bridge in your rack.
Book a technical briefing to scope your building, size the fleet, and model the 5-year saving against your current stack.