AI Service Layer

AI where the operator needs it

Horus's AI layer is built for hybrid deployment. When the ISP adds dedicated inference hardware in its datacenter, video streams can be processed close to the network and turned into new commercial features without replacing customer cameras.

Service-ready model catalog

• The model inventory can grow continuously as new use cases appear.
• Operators can enable packaged features for specific segments, sites, or accounts.
• New AI services can be monetized as soon as they are deployed in the ISP environment.

Choosing the right model

Each inference model analyzes video streams for a specific purpose: people or object detection, operational events, zone rules, vertical alerts, or custom use cases for a specific customer. Inference effort is not even or linear; it depends on the scope of the analysis, the number of cameras, stream resolution, and the technical capacity of the hardware installed at the ISP.

That is why Horus helps organize the catalog around expected return. The ISP can prioritize the models with the best ROI for its customer base, launch the highest-impact commercial features first, and scale inference capacity when demand justifies it.

From stream to event

• Horus consumes the customer's camera stream through the operator service layer.
• In the ISP datacenter, those streams are processed as inputs for inference engines running on dedicated hardware.
• Models generate events that are coordinated with Horus Cloud for notifications, mobile viewing, and operational workflows.
• The same service can add new models without replacing cameras or redesigning the field architecture.

Practical impact

By keeping inference in operator infrastructure, Horus reduces rollout friction, improves service responsiveness, and gives the ISP a fast path to differentiated offers based on analytics, detection, and event intelligence.