Research ยท Models
Models
Model families for orbital intelligence and distributed compute. Trained on public and computed data, evaluated on the held-out benchmarks, and exposed through the API and SDKs as they ship.
Orbital Intelligence
orbml-base
Base orbital-dynamics model for fine-tuning prediction tasks.
conjunctionnet
Conjunction probability with calibrated uncertainty.
maneuver-detect
Maneuver detection from TLE discontinuities.
satclass
Satellite behavior and type classification.
reentry-predict
Re-entry window prediction from decaying orbits.
Distributed Compute
gradient-compress
Gradient compression for high-latency space-ground links.
model-partition
Model partitioning across heterogeneous orbital nodes.
sync-scheduler
Synchronization scheduling under intermittent connectivity.
bandwidth-predict
Inter-satellite-link bandwidth prediction.
Access.
Models are reached through the CAE API and the Python/Rust SDKs; weights are released for the families that are open. Start with early access.