Satlyt, a space and AI software startup, raises $8 million to build software that runs AI models on satellites and coordinates computing tasks between them. Rather than manufacturing spacecraft, the company focuses on “orbital computing” services that can process data and help troubleshoot onboard issues in space.
According to reporting, Satlyt has completed two test missions. In one, it ran Google DeepMind’s Gemma model on a spacecraft operated by Momentus. Satlyt founder Rama Afullo says the test reduced error-related data transmissions by more than 60%, potentially lowering costs by avoiding repeated downlink and troubleshooting from orbit.
For its next mission, Satlyt plans to use a spacecraft from Indian startup TakeMe2Space. The mission involves multiple organizations: NASA is funding protocol testing in space, Stellerian is testing image processing for tracking objects in orbit, and TakeMe2Space is demonstrating third-party software hosting. Across outlets, Satlyt is described as aiming for open, reusable software compatible with satellites from different operators, with a longer-term goal of putting its software on a significant share of satellites over the decade.