A security camera generates far more information than a homeowner is ever likely to review. Add several cameras around a large property and the problem grows quickly, especially once AI begins analyzing the footage continuously rather than waiting for someone to open an app.
IC Realtime designed Bunker around the idea that much of that processing should happen closer to the cameras themselves. The new residential security platform runs on a dedicated gateway powered by NVIDIA Jetson technology, giving the system local computing capacity for video analysis instead of routing every core AI task through remote servers.
That architecture supports one of Bunker’s more ambitious features: learning what ordinary activity looks like at a specific property. The system analyzes recurring movement and timing so repeated behavior can gradually provide context for what happens next. A vehicle appearing at the same time each weekday can carry a different meaning from one arriving unexpectedly late at night, even though both events might look identical to a basic motion detector.
“On a large property, there is activity throughout the day, and a traditional system can treat all of it as equally important,” said IC Realtime CEO Matt Sailor. “Homeowners do not need more notifications or hours of footage to review. They need intelligence that understands the difference between normal activity and something that deserves their attention.”
Running that analysis at the property reduces the distance between seeing an event and interpreting it. Bunker can produce contextual alerts in plain language and maintain awareness across compatible camera feeds, while natural-language search allows someone to look for recorded activity by describing what they remember.
That last capability changes how archived surveillance video can be used. Searching for a particular vehicle or event has traditionally meant knowing roughly when it happened and working backward through footage. An indexed video system gives the user another way in, especially when the exact time is unclear.
Keeping More Intelligence Close to the Cameras
Video creates an obvious privacy question because the raw material can reveal far more about daily life than most connected-home data. Bunker does use cloud storage, but IC Realtime has placed the core AI processing on the dedicated edge gateway so routine analysis does not depend entirely on sending footage elsewhere before the system can understand it.
The company is careful not to describe edge and cloud processing as opposing choices. Its Standard plan includes 60 days of cloud video history, while Premium extends retention to 365 days. Remote mobile and web access remain part of the product as well, so cloud services still have a role even when more interpretation happens locally.
Sailor addressed that balance during CEDIA Expo in a session titled “Edge AI Cameras with Privacy,” where the discussion included the situations in which on-premises processing can improve response and data handling as well as the cases where cloud processing still makes sense.
For Bunker, local processing also supports continuous analysis. The intelligence does not have to be manually armed before it starts examining activity, which gives the software more of the routine behavior it needs to establish context. That matters because a system trying to recognize an unusual event first needs enough ordinary events to understand the difference.
Multi-camera awareness expands that context further. Instead of treating each camera as a separate source of alerts, Bunker is designed to interpret compatible feeds through the same intelligence layer. Activity can therefore be understood across the property rather than as a collection of disconnected detections.
The AI Layer Can Change Without Replacing the Cameras
IC Realtime has also separated the intelligence from the camera hardware itself. Bunker works with the company’s own equipment and compatible third-party devices, which gives integrators a path to add newer AI functions to systems already installed in a home.
That choice is especially relevant for large properties where replacing an existing camera network could turn a software upgrade into a major construction project. A dedicated gateway allows more of the intelligence to change independently from the devices collecting the video.
IC Realtime plans to keep updating the platform wirelessly after installation. Both subscription levels include automatic platform updates, so improvements to search or detection can be delivered without treating the original installation as the finished version of the product.
The model also reflects a larger change inside IC Realtime. The company has spent nearly 20 years building professional surveillance hardware and announced a $2 million Series A earlier this year to expand its AI and software work. Bunker puts those two parts of the business together by leaving cameras in their familiar role as sensors while moving more of the interpretation into software running nearby.
For residential video AI, that separation may matter more as models become more capable. Cameras already capture enormous amounts of information. Giving the system enough local computing power to understand more of that footage can improve response without requiring every new AI capability to arrive in the form of another camera mounted on the wall.
