According to reports, satellite software company Satlyt has completed an $8 million seed funding round, with plans to develop software capable of running AI models across multiple satellite platforms without building spacecraft themselves. The company previously deployed its software in two demonstration missions and successfully ran Google DeepMind’s Gemma model on a spacecraft operated by Momentus, reducing the volume of transmitted data related to onboard software errors by over 60%.

This week, Satlyt’s software will be launched aboard a spacecraft built by TakeMe2Space, with the mission involving NASA, Stellerian, and TakeMe2Space to test space cloud computing protocols, image processing tasks, and the capability of hosting third-party software on satellites. Satlyt expects to attempt connecting two satellites next year to establish a shared computing infrastructure; the company asserts its software can also enable on-orbit processing of sensor data or anomaly detection on spacecraft, thereby reducing the need to transmit data to ground stations.