The tasking brain above the marketplaces
Zerosat is a real optimization engine, not a lookup table. Every request runs on live orbital mechanics and a live forecast, and every decision is priced and explained.
Predict real access
Live orbital elements are propagated with SGP4 to compute exactly when each satellite can see your target, at what look angle, and at what achievable resolution. No revisit brochure averages, actual predicted passes for this request.
Turn weather into probability
Hourly cloud cover from the forecast grid is sampled over the exact target and pass window, not a daily city average, and each optical option carries the resulting usability number into scoring. Radar is scored on the same objective with weather risk near zero, so the engine knows precisely when a cloudy week makes SAR the rational buy instead of a gamble on a gap in the clouds.
Optimize one objective across vendors
Every feasible option from every vendor is scored on cost, delivery speed, resolution, and weather risk, weighted by mission priority. Competing requests for the same pass are arbitrated by priority, not first come first served.
Explain, deduplicate, record
The winner ships with a written rationale: what was chosen, what it beat, and why. Duplicate requests are caught before you pay twice for the same picture, and every decision, cost, and approval is recorded in a full audit trail.
One optimizer, every sensor
The same engine reasons across imaging and non-imaging phenomenologies, and chains them when one modality tips another.
Cross-cueing
An RF geolocation of an emitter can tip a confirming SAR collect, which tips an optical identification, which tips a hyperspectral look for material. The engine chains the modalities in time, so each confirming pass is bounded by the detection that cued it.
Phenomenologies
EO for detail, SAR to see through cloud and darkness, hyperspectral for material signatures, thermal for activity, and RF for geolocation of emitters. Each is scored on the same objective, so the pick is the best answer to the requirement, not the vendor a human happened to know.
Built to be trusted, not just fast
One adapter contract
Every vendor plugs in through a single adapter with STAC normalized metadata, so adding a constellation never changes the optimizer. Breadth grows without rewriting the decision logic.
Footprint-honest coverage
A polygon bigger than one scene is split into scene scale cells and planned as a mosaic. A sensor whose swath cannot span a cell is never credited with covering it, so a plan never promises pixels it cannot deliver.
Explainable by construction
The rationale is not a marketing afterthought, it is how the optimizer selects. Scoring math, the alternatives considered, and the winning tradeoff are recorded for every decision.