A software and analysis layer above the sensor market
We produce interpreted force posture intelligence, built from data ingested through third-party providers and sovereign constellations.
Coverage can be procured. Correlation and theatre level context cannot
Sensor types are bought separately, delivered in different formats and revisit cycles, and read by different teams. A ministry with a SAR feed and an RF feed has two data streams, not one assessment.
Each additional stream adds to the analytical load it was bought to relieve. Correlation and context are the parts that do not scale by buying more data, and they are the parts we are built for.
We are most useful when the subject is trying to hide
Concealment tends to work against one sensor at a time. Cover defeats optical, emission discipline defeats RF, and dispersal defeats a single revisit. Defeating several sensors at once, across a network of facilities, over months, is considerably harder.
We hold what one sensor does not show against what the others do, and surface the contradiction. A facility that has gone quieter than its own history predicts is as interesting to us as one that has become busier.
Three layers
We normalise formats and align observations geometrically, radiometrically, and in time, so that a difference between two of them is a real difference and not an artefact of how each was captured.
Machine learning models establish what normal looks like at a specific facility from years of its own history, allowing for seasonal and operational variation. Classical change detection and statistical scoring run alongside them, so every score can be traced back to the observations behind it and defended in an assessment.
Scored detections are assembled into the connected account an analyst would otherwise build by hand, with the timing, the consistency of type class, and the confidence stated. It is written to be read and challenged in an assessment, not cleared as a queue.
What the platform does today
We monitor named facilities using SAR and optical data, scoring activity against multi-year baselines and ranking pre-deployment indicators. Force concentration work extends this to RF and thermal data, correlating departures at one facility with arrivals at another.
Where we are going
Detection across open terrain and work in the maritime domain are next on our roadmap. Both extend the same baseline approach beyond fixed facilities to movement in the field and at sea.
Strategic assessment, in the indicators and warnings function
Our users are intelligence staff and programme planners forming judgements about force posture and intent, on the timescale of a warning rather than an engagement. The platform gives them a ranked, reasoned account of what has changed across a facility network, and the confidence attached to it.
This is not about nominating targets. It is about understanding how the adversary behaves, which is why we sit upstream of the targeting cycle rather than inside it.
Your instance, your baselines, your assessments
We deploy into a sovereign instance you control, ingesting data under your own commercial arrangements. Processing, baselines, and finished assessments remain inside it.
The layer where data becomes intelligence is the one that decides who can be pressured, delayed, or cut off. Holding it inside your own instance means the baselines you build and the judgements you draw from them remain yours, and are visible to nobody else.