Active contacts –
– UTC
T+0:00
A satellite is out of contact with the ground for about 74% of every orbit. Orbtrix is building the AI that makes decisions during that time, and the constellation that runs it.
A single satellite cannot carry the compute this needs, so we move it onto dedicated compute satellites. Each node is built around its GPUs, with power, thermal and radiation design done once on our platform. The customer satellite adds only a compact Ka-band terminal with an antenna of about 10 cm, linking to the nearest node at a design target of 1 Gb/s; nodes connect to each other over an optical mesh at 2.5 to 10 Gbps.
Through a single API the customer satellite calls Neeli, our onboard autonomy model, for diagnosis, replanning and tasking in seconds to minutes, or deploys and trains its own models on the constellation and calls them the same way. A lightweight agent on the customer satellite handles safety and fault isolation in milliseconds, and DISHA on the ground plans the fleet, trains models and keeps oversight of any action outside agreed limits.
A first-pass constellation study sized the system from orbit geometry: a full-year simulation of five sample customers in mixed low Earth orbits, scored on the worst-served customer at the worst time of year. 23,502 layouts were screened, the selected designs were checked continuously for 365 days, and results were cross-checked by numerical integration to within 0.05 percentage points.
The result is a Walker constellation at 800 km with eight nodes per plane, inside optical link range and below the inner radiation belt. 24 nodes give a first service with no customer out of contact for more than 20 minutes. 32 nodes bring the longest gap under 10 minutes. 48 nodes give near-continuous supervision, with every sample customer linked more than 99.98% of the year and no gap longer than one minute.
The two largest design drivers are the customer antenna, which needs arrays facing four directions, and a link that closes at about 3,500 km. Confirming that link budget and the terminal design are the next steps.
Worst customer and average share of the year linked. Five sample customers, full year.
Longest gap in a year. Minutes, log scale.
For more details, contact us at info@orbtrix.space
Project E is shared compute and connectivity in orbit. Anyone whose software has to run close to a satellite can use it: operators who fly satellites, and companies that build software, models and analytics for them.
| Customer | What they need | How they use Project E |
|---|---|---|
| Satellite operatorsEO, SAR, communications fleets | Decisions while their satellites are out of ground contact; fewer operators per satellite. | Fit the Ka terminal and API client. Neeli diagnoses, replans and tasks their satellites in seconds to minutes, under DISHA oversight. |
| Space software (SaaS) companiesNew segment | A place to run their product next to customer satellites, without building hardware. | Deploy their software on Project E nodes and sell it to satellite operators through the same API. |
| AI companiesPerception and navigation algorithms | A way to get models onto satellites without each operator integrating them on board. | Host models on our nodes. Customer satellites call them directly, so one deployment reaches every connected satellite. |
| Defence operatorsGovernment and allied fleets | Secure, resilient compute and storage that keeps working under interference. | In-orbit storage of models and cryptographic keys, signed commands, a jam-resistant optical mesh and isolated compute. |
| Analytics companiesImagery and data analytics | Faster results from imagery bought from imaging vendors. | Process images in orbit as they come off the vendor's satellite, and send results to the ground instead of raw data. |
Dates are Orbtrix targets. Results are from the mission study: worst of 5 sample customers over a full year, customer Ka arrays on 4 sides, 3,500 km link. Orbits for Q4 2027 and Q3 2028 are illustrative; ground stations are assumed.