Demo
Sample workflow.
Interactive demo
Orbital Workload Planner
Adjust mission parameters—outputs update instantly. Commit to persist a snapshot in the compute console.
Downlink reduction
79%
~1.6 TB/day edge-filtered vs raw 2.0 TB/day
Bottleneck score
100/100
Post-filter egress
~13 TB/mo
Orbital compute plan
- ↗Stage Object Detection on SmallSat Bus with 32 TOPS budget.
- ↗Target 79% downlink reduction vs raw 2.0 TB/day sensor-adjacent load.
- ↗Align scheduler to Balanced priorities with seconds latency class.
- ↗Reserve 0.45 TB/day link-shaped egress after edge filtering.
Workload routing graph
util 67%Inference schedule
- T1Burst inference after each capture sequence with thermal cooldown gaps.
- T2Policy router elevates tier-1 events ahead of calibration batches.
- T3Eclipse-side batch compaction for low-priority archive streams.
Model deployment timeline
Artifact ingest
complete12 min window · stage 1
Bus compatibility matrix
complete28 min window · stage 2
Shadow inference
active45 min window · stage 3
Fleet promotion
queued60 min window · stage 4
Mission data priority queue
- #118%
High-confidence detections
immediate promotion
- #212%
Thermal excursions
rolling 3-orbit window
- #322%
SAR change chips
diff against last downlink
- #48%
Calibration ladders
batched nightly
- #540%
Full-frame archive
ground recall only
Cost & fit
Modeled $85.3k / month downlink avoidance · Mission fit 78/100 · Runtime 364h edge/mo
For SmallSat Bus streaming Optical Imagery at ~2.0 TB/day (adjacent), SpaceCompute Cloud plans Object Detection on SC-Edge Compact. Expect ~79% less data on the wire, ~$85.3k/month avoided downlink spend at modeled rates, and mission-fit 78/100.