01 / Thermosphere
AKASX Drag
0%
storm-time density error · vs 13.11 % best CCMC model · NASA/CCMC CAMEL
Measured with a look-ahead NRLMSIS input; real-time results vary.
AKASX
Physical AI×Space AI
The sky, measured.
LEO · 400–600 kmT+00:00:00F10.7 · Kp · Dst
Scroll
02 / Manoeuvre
AKASX Custody
0km
median position error after a burn · vs 3.14 km SGP4 on the last TLE
03 / Sun
AKASX F10.7
0sfu
1–3 day solar flux error · vs 10.50 sfu SWPC
04 / Knowledge
AKASX Science Search
Air‑gapped.
space-science literature search, fully offline
05 / Device
AKASX Edge
Any CPU.
the models offline · licensed · self-testing
06 / Pole star
AKASX DHRUV
Cites its sources.
the analyst co-pilot · cites its sources
Thesis
Physical AI×Space AI
Models of real physics. Atmosphere, orbit, sun. Measured, then shipped.
What the models can do
Measured where it works.
Honest where it doesn't yet.
Live · measuredProvenIn developmentNext
Drag
9.53 %Storm-time density error vs 13.11 % best CCMC
Custody
1.58 kmPosition after a manoeuvre vs 3.14 km SGP4
F10.7
8.47 sfu1–3 day solar flux vs 10.50 SWPC
Science Search
0.966 MRRFinds the right space-science data, offline
DHRUV
0.0 %Cites sources · fabricated orbital numbers
Edge
Any CPUAll models offline, any CPU, self-checking
Manoeuvre detection
AUC 0.84Who moved, across the catalogue
Burn scars
IoU 0.81Wildfire mapping vs 0.729 U-Net
Lifetime & re-entry
How long it lives, where it comes down
Collision inputs
Better density and custody for Pc
Storm forecast
Kp and storm warnings, hours ahead
Telemetry health
Anomalies before they become failures
Covariance
How sure every position is
Onboard drag
Storm-aware satellites, no fresh uplink
Onboard triage
Decide in orbit what to downlink
Ionosphere · HF · radiation
Impacts on grids, aviation, GNSS
Mission
Make spaceintelligently predictable.
Today on the ground. Next, on board.
Every kilometre · every percent
- Search time
- Imaging time
- Fuel
- Mission life
Who it is for
Film 02