Fusion & Command
Most operations centers have a wall of screens because each system arrived with its own console. Correlation happens in a person’s head, under time pressure, using whichever windows they happened to be looking at.
Mentat builds the layer underneath that wall: a common event model, a shared spatial and temporal index, and correlation rules that surface related signals together rather than as four independent alerts arriving in four different places.
The engineering problem is unglamorous and decisive. It is schema reconciliation, clock synchronization, geospatial joins, backpressure handling, and the discipline to keep latency bounded when every upstream system has different failure characteristics.
What we build
One event model that every source maps into, so a camera detection, an LPR read, and a sensor alert can actually be compared and correlated.
The value of a fusion layer is measured by how much noise it removes. Four related signals should produce one alert, not four.
A fusion layer becomes critical infrastructure the moment operators trust it. It has to degrade predictably rather than fail completely.
What the system showed, when it showed it, who saw it, and what they did next, captured as a reconstructable record.
How we get there
One canonical representation that every feed maps into, designed with the queries and correlations you actually need in mind.
Per-system adapters translate vendor formats into the canonical model, isolating the rest of the platform from vendor change.
Spatial, temporal, and rule-based correlation collapse related signals into single actionable items with preserved provenance.
Latency, alert volume, acknowledgement time, and false-alert rate measured continuously so the system can be tuned against reality.
Governance & safeguards
Combining feeds creates a more powerful capability than any single system, and that is exactly why the controls belong at the fusion layer rather than being left to each source system to handle differently.
One authorization model across all fused sources, so combining feeds cannot become a route around a restriction on any individual feed.
Every fused item retains which source produced it, at what confidence, and when, so a correlation can always be traced back to evidence.
Correlation surfaces information for operators to act on. Consequential actions require a person, and the system records who that was.
Built on
We build on established open-source and cloud-native building blocks rather than proprietary black boxes, so the platform stays inspectable, portable, and yours.
Streaming
Correlation
Reliability
Related
Too many consoles, not enough picture?
We can map your current sources, model the correlation you actually need, and design the fusion layer that reduces alert load instead of adding to it.