Earth observation
Start with public Sentinel and Landsat observations. Source selection depends on location, season, cloud cover and the scale of the feature being studied.
THE ENGINE / TECHNOLOGY
The intelligence between orbit and action.
Earth observation gives us the pixels. Geospatial engineering makes them comparable. AI helps reveal what they mean.
01 / THE PERSPECTIVE
From orbit.
To answer.
THE PURNOVA INTELLIGENCE LAYER
There is no shortage of satellite data.
There is a shortage of clarity.
We build the intelligence in between.
◈ SIGNAL PATH / EXPLORE THE PROCESSCONCEPTUAL WORKFLOW
SATELLITE SOURCES⟶MULTISPECTRAL IMAGERY
Start with an area of interest. Select suitable observations from existing missions, with acquisition date and source provenance attached.
01/ 04EXISTING MISSIONS.
NEW POSSIBILITIES.Copernicus Sentinel ↗USGS / NASA Landsat ↗
NEW POSSIBILITIES.Copernicus Sentinel ↗USGS / NASA Landsat ↗
Independent intelligence.
No satellite ownership required.
SYSTEM ARCHITECTURE
Built around the signal.
Designed for context.
Our approach and platform design.
Features under development are identified explicitly.
EO sources →Ingestion →Normalization →Tile processing →AI / computer vision →Change engine →Spatial database →Reports / planned API →
Geospatial processing
Align coordinate reference systems, clip an area of interest, mask unusable pixels and normalize comparable imagery before analysis.
AI / computer vision
Use segmentation to separate land-cover classes and detection to identify visible features. Models must be validated against the intended geography and imagery resolution.
Change detection
Compare aligned observations. Differences become candidate events, with image quality and seasonal effects considered before interpretation.
Temporal intelligence
Look across multiple dates. Persistent trends, seasonal cycles and isolated anomalies tell different stories and require different baselines.
Data infrastructure
The architecture uses imagery object storage, spatial metadata and queued processing jobs. Separate ingestion from inference so work can be retried and traced.
APIs & automation
Planned delivery includes maps, PDF reports and API integration. The self-serve platform and public API remain in development.
Security by design
Platform design priorities include scoped access, protected customer boundaries, encrypted transport and auditability. Specific controls will be agreed for each deployment; no certification is claimed.
Scalability & review
Tile-based processing and asynchronous jobs support larger regions. Human review and explicit quality checks remain part of responsible interpretation.
APPLIED AI
A method for
every question.
We begin with established techniques and develop domain-specific models as pilot data and validation allow.
METHOD EXPLORER / CONCEPTUAL VISUALIZATION
Computer vision
A spatial signal becomes useful only when interpreted with source quality, temporal context and domain knowledge.
Illustrates a method, not measured model performance.A WORKFLOW, EXPRESSED
Define the region.
Ask a better question.
A conceptual view of an API-driven workflow. This is illustrative pseudocode, not a released SDK.
Explore the platform roadmap ↗CONCEPTUAL CODE / NOT A PUBLIC SDK
observation = purnova.observe(
region=farm_boundary,
interval="weekly",
analysis=[
"change_detection",
"crop_health"
]
)
insights = observation.analyze()Resolution is a constraint.
Context is essential.
Public imagery is valuable for regional patterns, but small structures may require licensed higher-resolution data. Cloud cover, acquisition timing and seasonal change affect usable observations. A pilot establishes what is observable before defining delivery frequency.