Catalyst.earth Logo
Catalyst.earth Logo

CATALYST OnBoard Edge · New

Analysis-ready imagery,
before it reaches the ground.

CATALYST OnBoard Edge runs the full processing pipeline on satellite hardware, producing orthorectified, analysis-ready imagery before a single byte is downlinked.

A satellite flying above South East AsiaDiscuss your architecture
Today, satellite data is downlinked before it can be processed, consuming bandwidth, ground infrastructure, and time. CATALYST OnBoard Edge changes the order of operations, running orthorectification, band-to-band alignment, and AI analytics on satellite hardware so only the data worth moving reaches the ground, already analysis-ready in near real time. It is the same 45 years of photogrammetric science that runs in CATALYST OnBoard Cloud, now on the satellite.
The benchmark
0 s
to a full processed multispectral scene

A full multispectral scene, complete orthorectification and band-to-band alignment, processed on satellite hardware before a single byte reaches the ground station.

Edge is designed to run across a range of processors, scaling from low-power to high-performance configurations to match your mission constraints.

Processing benchmark measured on NVIDIA Jetson Thor. CATALYST OnBoard Edge is a validated capability demonstrated in benchmark; it is not yet in operational deployment. Figures are representative of the tested configuration.

Multispectral

full scene processed on orbit

4.75 m

ground sample distance

1.2 GB

raw scene processed on orbit

L1A → L1C

full processing level

0.96 F1

AI ship-detection accuracy at the edge

15W+

scales from low-power to high-performance hardware
CATALYST OnBoard Edge

Move the processing, not the raw data

Everything CATALYST OnBoard does in the cloud, running on the satellite, so intelligence is ready the moment it lands.

Process before downlink

Produce orthorectified, analysis-ready imagery on the satellite. Process and run AI on the satellite so an actionable result is ready during the pass, cutting the time from capture to decision from hours to seconds.

Compatible with every NVIDIA Orin

Runs across the full NVIDIA Orin family, from low-power Orin NX up to high-performance NVIDIA Jetson Thor. Match the module to your mission's size, weight, and power budget.

CPU pipeline, GPU free for AI

The processing pipeline runs across CPU cores, leaving the GPU and Tensor core free to run AI inference in parallel.

On-orbit AI analytics

Ship detection, wildfire and flood mapping, cloud masking and more run at the edge, so alerts are generated before the pass is even over.

Less to downlink

Discard cloud-covered or unusable scenes before they use bandwidth, send compact detections and alerts in place of full scenes, and prioritize what comes down first. Lower downlink volume can mean a smaller, less costly ground segment.

Same platform as the cloud

Cloud and Edge run the same containerized algorithms, on the ground or in orbit. Build once, deploy anywhere.
When minutes matter
Processing that keeps pace with the sensor
A satellite captures an image strip in about 13 seconds. CATALYST OnBoard Edge now processes that strip  in 5.5 seconds, faster than the satellite captures the next one.  Analysis-ready imagery and AI detections are available within seconds of capture, not hours after downlink.
Satellite captures a strip
~13s
Edge processing today
5.5s
On a path toward
<5s
TARGET
For missions where minutes decide the outcome, such as maritime interdiction, disaster response, or threat detection, that is the difference between acting on what is happening now and reacting to what already happened.
On-orbit AI analytics
Intelligence, not just imagery
Because the processing pipeline runs on the CPU, the GPU and Tensor core stay free to run AI inference in parallel to get actionable intelligence faster.

CATALYST has demonstrated a range of AI use cases at the edge, turning raw captures into alerts and detections before the data is downlinked. Many are built for defence and government missions such as maritime surveillance, threat detection, and disaster response.
  • Ship & vessel detection
  • Land & border detection
  • Autonomous tip & cue
  • Flood extent mapping
  • Volcano monitoring
  • Cloud masking
  • Vegetation indices
  • Wildfire detection
Architecture
Built to run two jobs on one board
The same Docker technology and Python SDK that power the cloud deployment run at the edge. The processing pipeline is split from the AI workload by design.
Failsafe by design: raw data is written to disk before processing begins, so an unprocessed capture is never at risk.
CPU
Processing pipeline
Band alignment, georeferencing and orthorectification run across CPU cores.
GPU
AI analytics
The GPU and Tensor core are reserved to run AI models in parallel.
CPU and GPU, choreographed for maximum performance.
TRUST
Designed to be trusted on-orbit
1
Combine edge and cloud
Run OnBoard Edge for AI answers in seconds to drive tip and cue, and still downlink the full scene to OnBoard Cloud for standard processing and archiving. Task-optimized algorithms let you apply the right processing to each job.
2
The science is not new
Edge deployment runs the same validated pipeline that processes hundreds of millions of scenes on the ground today. The output is the result you already trust, produced by the same algorithms.
3
You stay in control
Your engineers configure the modules to match what the mission needs. It is not a black box you hand your data to.
Who it is for

Two missions, one capability.

Defence & mission-critical

When intelligence is time-sensitive, the value is not compute saved: it is how fast an actionable detection reaches the people who need it.

Processing and detecting on orbit collapses the gap between capture and decision.
  • Detections generated at the time of capture, not after downlink
  • Maritime surveillance, disaster response, change monitoring
  • Autonomous tip and cue to retask assets faster

Commercial constellation operators

Edge extends the proven CATALYST pipeline onto your satellite.

Move only the data worth moving and lighten the load on your ground segment.
  • Lower downlink volume can reduce ground-station cost
  • Inherited photogrammetric accuracy, no science to build
  • Relay satellite-to-satellite and direct to device
  • The same pipeline as cloud deployments, carried to the edge with no rewrite
Design partner program
Shape your edge configuration before you launch
We are working with a small number of operators to bring CATALYST OnBoard Edge to orbit.
As a design partner we collaboratively define the workflows, hardware target, and performance envelope for your mission and you get early access to the pipeline as it moves from validated benchmark toward operational deployment. 
If you are planning a mission where on-orbit processing matters, the time to talk is now.
Get in touch
Bringing Edge to your constellation?
Tell us about your mission and hardware. We will walk you through what CATALYST OnBoard Edge looks like for your architecture.
  • Full processing pipeline on satellite-class hardware
  • On-orbit AI analytics on the same compute hardware
  • Hardware-agnostic, from low-power to high-performance
  • The same proven science as cloud deployment
Talk to our team
Fill this in and we will be in touch within one business day.

"*" indicates required fields

Your Name*
Join our newsletter?

Hidden Fields

Fields that are used to capture dynamic parameters for tracking and marketing purposes.

envelopephone-handsetcross