Spica Space
EO / Boulder
Mission nominal
Spica Insight fuses an 11-satellite sub-meter constellation with a proprietary on-orbit AI stack, returning decision-grade change detection within 90 minutes of acquisition. Authenticate once, task a satellite, receive analyzed COG, Zarr or GeoParquet tiles via webhook — at a median tasking-to-delivery latency of 78 minutes.
Spica Insight is a single auditable pipeline. Every stage is timestamped, every artifact is versioned, and every result lands in your stack through a signed webhook — so the difference between "the satellite saw it" and "your system acted on it" collapses to a measurable SLA.
An authenticated POST /v1/tasking with an AOI, a window and a product class. The scheduler returns a pass ID and a confidence-weighted acquisition window within 800 ms.
The next satellite in the 11-vehicle constellation overflies the AOI at 0.7 m native resolution, with a 4-hour median revisit across 94% of Earth's landmass.
A radiation-hardened inference engine runs change detection, cloud masking and atmospheric correction in orbit — reducing downlink volume by 22× versus raw raster shipping, cutting customer bandwidth costs per our 2024 benchmarks.
Downlinked chips are orthorectified, tiled and packaged as Cloud-Optimized GeoTIFFs, Zarr cubes or GeoParquet — then routed through the Boulder, Toulouse or Singapore ground station closest to the customer.
A signed webhook fires with the tile URIs, quality flags and a confidence score. Median end-to-end latency: 78 minutes. Independently benchmarked by the Open Geospatial Consortium in Q1 2025.
Compliance posture. SOC 2 Type II, ISO 27001 and FedRAMP Moderate (the sole EO provider authorized at that level as of November 2024). Data residency selectable per AOI: US-only, EU-only or APAC-only processing paths.
No proprietary GUI required. The platform speaks HTTP, streams webhooks, and ships tiles in the three formats your stack already loads.
A flat, RESTful surface with predictable JSON. Idempotent tasking, paginated history, signed S3-compatible tile URLs and a first-class /v1/changes endpoint that streams only what changed since your cursor.
POST /v1/tasking — submit an AOI and windowGET /v1/tasking/{id} — poll acquisition stateGET /v1/changes?since= — diff against your last cursorGET /v1/tiles/{id}/manifest — signed URL setEvery product ships in all three: Cloud-Optimized GeoTIFF for raster pipelines, Zarr for cube-native analysis, GeoParquet for vector change events with embedded confidence and provenance columns.
HMAC-signed POST callbacks with exponential backoff, a 7-day replay window and a dead-letter inspector in the dashboard. Push to S3, GCS, Azure Blob or any HTTPS endpoint you control.
Official spica-python and spica-ts clients with full type coverage, plus a spica CLI for shell-driven tasking and bulk backfills. Plus the open-sourced spica-change PyTorch library (cited in 140+ peer-reviewed papers) for in-house model fine-tuning.
A web console for AOIs, tasking history, spend and access control — useful for the GIS lead who wants answers without writing curl.
Every tile carries sensor ID, acquisition time to the millisecond, processing chain version and model commit SHA — so a regulator-facing report can trace any pixel back to a specific satellite pass.
A four-step onboarding that produces a verifiable webhook in your terminal before lunch. No pre-commit, no annual contract, no proprietary GUI to learn first.
Generate a scoped API key in the dashboard or via POST /v1/keys. Keys inherit RBAC roles — read-only, tasking, billing — so the same key never crosses a privilege boundary.
POST an AOI as GeoJSON, pick a product (change-v3, flood-extent, methane-plume, crop-loss) and a window. The response includes the pass ID and a confidence-weighted ETA.
Register an HTTPS endpoint or an S3/GCS/Azure Blob bucket. Spica POSTs a signed manifest the moment processing completes — your service acks, tiles download, and your ETL kicks off without polling.
push, no pollingPull the Cloud-Optimized GeoTIFF (or Zarr / GeoParquet) from the signed URL and load it with rasterio, xarray or DuckDB. No proprietary reader, no license server, no offline activation.
≈ 78 min median end-to-endSandbox keys are free, throttle at 5 taskings per hour and draw on the same public datasets released through our Microsoft Planetary Computer partnership — 3.8 million km² of tropical forest, ready for testing on day one.
The contract is deliberately boring. Predictable JSON, predictable status codes, a single signed payload on the way back. If you can read a REST API, you can integrate Spica Insight without reading another PDF.
# tasking.py — submit an AOI for change detection
from spica import Client
client = Client(api_key="sk_live_...")
task = client.tasking.submit(
aoi="aoi/colorado_mine.geojson",
product="change-v3",
window={
"start": "2024-11-08T00:00Z",
"end": "2024-11-09T00:00Z",
},
min_confidence=0.85,
)
print(task.id, task.eta_utc)
# → task_8f3c1a 2024-11-08T14:22Z{
"event": "tasking.completed",
"task_id": "task_8f3c1a",
"acquired_at": "2024-11-08T13:48:11Z",
"delivered_at": "2024-11-08T14:22:07Z",
"latency_seconds": 2036,
"sensor": "SP-07",
"resolution_m": 0.7,
"tiles": [
{
"format": "cog",
"url": "https://dl.spica...",
"checksum": "sha256:9f3a..."
},
{
"format": "geoparquet",
"url": "https://dl.spica...",
"changes": 147
}
],
"signature": "hmac-sha256=4e7b..."
}Both the tasking client and the webhook verifier are open source under MIT — see the spica-python repo on GitHub. Test against the sandbox with the same code path you ship to production.
We will task the constellation against your real AOI during the call, walk you through the resulting webhook, and answer the integration questions your platform team is already drafting. 380+ organizations — including a leading global reinsurer and the European Environment Agency — run Spica Insight in production; the demo is where you find out if the fit is mutual.