Compare two AI models in orbit.
A research team needs to understand whether its detector performs well on flight hardware, within the available runtime and resource limits.
A hypothetical brief to explain the service. This is not a completed customer project or a confirmed offer.
The request
- 2 compatible model variants
- 100 agreed, licensed input images
- One operator-approved payload computer
What comes back
- Prediction outputs linked to each input
- Inference latency and available resource telemetry
- Execution logs and model-version identifiers
What needs to be confirmed?
Model format, CPU or accelerator support, memory, data rights and access to telemetry must be confirmed. Model files are supplied through an agreed secure transfer after review.
What does it cost?
Technical quote required. Integration, validation, onboard execution and return of the agreed results must all be included.
What changes compared with launching your own satellite?
A compatible payload already in orbit removes the need for a dedicated spacecraft and launch. A short campaign may be possible within weeks after acceptance; integration and available operational windows determine the actual schedule.
| Dedicated satellite | Existing compatible capacity |
|---|---|
| Develop and qualify a spacecraft | Review and integrate the workload |
| Procure and schedule a launch | Agree an available operating window |
| Operate the entire mission | Buy the scoped task and its results |
This compares project scope, not guaranteed cost or time savings. A hosted workload still requires technical acceptance and an operator agreement.