Calibrated answers, wherever the machine is.
Semaphore trains physical AI models on its digital worlds. Machines get their answers three ways.
01 cloud
Any connected machine calls our models through software interfaces.
02 broadcast
Our orbiters and rovers run the models and broadcast answers to machines in range.
03 onboard
Customers run our models on their own vehicles, with no link needed.
Why it matters
A rover driven from Earth waits on every command. NASA planned ten drivers in three shifts for its Volatiles Investigating Polar Exploration Rover (VIPER). With those waits, VIPER was to average about 1 cm per second, a twentieth of its top speed. Answers from our orbiters or from onboard models cut that wait. (NASA on VIPER operations; NASA Ames on VIPER traverses)
Trained and tested
We train our models on our digital worlds and test them on data they never trained on. Before a model informs an operation, we assess its credibility under NASA's standard for models and simulations. (NASA on models and simulations)
Citation of sources such as NASA does not imply their endorsement.