Satellites that decide in orbit
Ouranos builds ultra-low-power physical reservoir computing on FPGAs, so a spacecraft can sense, judge and flag in orbit, and send home only what matters.
Listening costs power
A satellite listens to every sensor, spends up to 15% of its power on cooling, and sends home about 2% of what it sees.
of a satellite's power can go to thermal control alone, before any computing is added.
of a 3U CubeSat's power budget goes to a 1.5 W flight computer.
of observed imagery is ever downloaded. Problems surface only after the data reaches Earth.
Most small satellites judge their data only after it reaches Earth.
Like a doorbell camera that streams every second of video to a distant office before telling you someone is at the door. Ouranos puts the judgement on board.
A reflex, not a brain
Your hand leaves a hot plate before you think. Ouranos gives a satellite the same reflex.
A reservoir on FPGA watches the sensors and wakes the main computer only when something changes, so the satellite stops spending power on listening.
1Input signal
Sensor data such as imagery, vibration or RF streams in as a raw signal.
2Reservoir
A fixed, high-dimensional circuit on the FPGA responds. It is never trained or reprogrammed.
3Readout
A small trained layer reads the reservoir's state and outputs a decision, onboard.
Inside the reservoir, a signal ripples through fixed connections, like a stone dropped in water. The ripples already carry what the signal means, so only a small readout has to learn anything.
Try the reflex
Reservoir watching
An illustration, simulated in your browser. On the bench, the reservoir caught every temperature spike in our labelled test set.
What it means for a mission
Real-time
Decisions happen in orbit, without waiting on a ground pass to know what matters.
Low power
Designed to fit the tight power budget of a small satellite.
Reconfigurable hardware
Built on FPGA, so the same chip can be reprogrammed for new sensors or missions.
Affordable
Built on hardware the space industry already trusts, keeping development and deployment costs down.
The road to orbit
Prototype the reservoir
Validate the assumptions
Done: TRL 4Find a test-bed partner
Start proof-of-concept tests
NowFirst MVP
Test to industry needs
PlannedFirst product
First customer in the field
PlannedProduct V2
Ready to raise capital
PlannedSpace-qualified design
Uses beyond space
Planned
Our ask today: a satellite team that wants to run a proof-of-concept test with us, and investors who want an early look at where this goes.
Connect with usFour people, four countries
Who we are
A deep-tech computing company developing ultra-low-power AI for extreme environments, starting with space.
Who we are not
Not another general-purpose AI processor, and not an FPGA, chip or satellite maker. One computing architecture, aimed at one problem.
- AE
Ariel EcheverryCo-founder, CEOColombiaGeopolitics, trade and the space economy.
- MN
Mohammad NatshehCo-founder, AerospaceJordanSystems engineering for the hardware.
- HS
Harsh ShahCo-founder, OperationsCanadaMechanical engineering and product commissioning.
- RA
Rafael AlzateAdvisorSingaporeStart-up acceleration and guidance.
Connect with us
Building satellites, working on on-board AI, or curious about what we do? We would like to hear from you.
We especially want to hear from
- University and industry satellite teams
- CubeSat and IoT-satellite makers
- Low Earth orbit operators
- Partners and investors in deep tech
Write to us
a.echeverry@ouranos.coTell us what you are building and what you need.
Send an email