Solutions
Our approach
We close the gap between what was expected and what is happening.
Every satellite behaves differently. The same design, the same batch, the same orbit, and still each one drifts its own way. So nothing written before launch stays true for long: procedures written for the class meet spacecraft that each behave slightly differently, and failure rates from the datasheet meet components that were actually flown a specific way.
Keeping that reconciliation current is per-satellite work, continuously, which is why it does not get kept up by hand. Our approach is to model the individual satellite rather than the class, and to do it on the telemetry it already sends down.
That is what keeps both cases live: the operational case you wrote before launch, and the reliability case you signed off at design review.
Assurance is not detection
Detection asks whether something changed. Assurance asks whether the fleet is still being operated as intended, and what to do when it is not.
Answering that continuously takes three things: knowing what "as intended" means for this fleet, noticing the ways reality departs from it, and producing something a person can act on and defend afterwards. Detection is one part of the second. That is the whole difference between a tool that tells you something moved and one that tells you whether it matters.
We model accumulation, not just change
Anything you can watch works on change. But the most expensive decisions in a fleet are not about change, they are about accumulation. A component that has absorbed years of thermal cycling and deep discharge reads nominal on every check, right up until it does not.
Nothing you can watch will tell you that. It has to be modelled.
We model how each component ages under the stress it actually experienced, and forecast where its reliability is heading. End of life stops being an assumption and becomes evidence.
Two cases, one architecture: the operational case kept live, and the reliability case kept live.
We learn each satellite’s own normal, then decide what matters
We measured our detector against 14 years of real mission data using the ESA Anomaly Detection Benchmark. It reached a ROC-AUC of 0.936, a competitive result on a public benchmark. And yet, even at that level, standalone detection generates a continuous stream of false escalations that land on operators as uninvestigated noise.
Most tools help you see more. We help you look at less.
Deterministic reasoning, informed by AI scoring
Our approach combines advanced deterministic logic with AI scoring, using mission context, spacecraft state, scheduled activities and cross-parameter conditions, to determine what actually requires attention.
What reaches the operator is not a score. It is a decision-ready event: context attached, history surfaced, response suggested, authority preserved.
All of it runs alongside the mission control system you already use. No re-architecture, no replacement of existing tools, no change to how you fly.
Every number traces to a cause
The models are deliberately transparent. They collapse to your qualified design baseline when the vehicle is flown as designed, and diverge only as measured behaviour departs from it. Every parameter is exposed and every default traces to a recognised standard or to in-orbit calibration. No hidden constants, and nothing you cannot defend in a review.
Why this is not a weekend project
Most of what we do can be built. Teams do build it: a dashboard here, a script there, a rule written after the last incident. What is hard is not the build, it is that the reconciliation is per-satellite, continuous, and never finished. Procedures change, orbits decay, components age, and people leave. Homegrown tools are good enough until the fleet doubles, because they were built for the fleet you had.
The maintenance is the product.
The workload grows even if the fleet does not
Collision risk in low Earth orbit keeps rising, and conjunction alerts now arrive in the hundreds per satellite per week, with a handful each week requiring real analysis. That work is set by the environment, not by your fleet plan. Every year you fly the same satellites, they cost more to fly.
The same team. More spacecraft. Decisions they can defend.