Satlyt Inc., a startup building AI software for satellites, has raised $8 million in seed funding led by early-stage fund Non Sibi Ventures, SiliconANGLE reported. Satlyt was founded in 2024 by Chief Executive Rama Afullo, who previously worked as a product manager on SpaceX's Starlink business and earlier as a cloud architect at Google.
Most satellites carry limited onboard computing and instead send the data they collect down to ground-based servers for processing. That approach is constrained because the average satellite is only within range of its operator's terrestrial antenna for a few minutes each day, limiting how much data can be transmitted, and the data loses some accuracy on its way down. Satlyt's platform is designed to let satellites equipped with graphics processing units process some of that data internally, reducing the volume of information that must be sent to Earth.
The first satellite to carry Satlyt's software launched into low-Earth orbit in March, built by venture-backed startup Momentus Inc. and running Google's open-source Gemma 3 1B language model. In a test that simulated common satellite software errors, the platform used Gemini to collect error logs, identify root causes and suggest fixes, cutting the amount of telemetry that had to be beamed down to Earth for analysis by more than 64%. As of April, Satlyt was working to upgrade its stack to a newer Gemma 4 E2B model and aiming to reduce that model's memory footprint by more than 90%.
Another Satlyt-powered satellite was scheduled to launch October 1 aboard a SpaceX rocket. NASA plans to use the company's software to test new orbital computing methods, while space-observation startup Stellerian Inc. intends to run image-processing algorithms aboard the same system. Satlyt is reportedly also developing a tool to distribute computing workloads across multiple satellites, with the first system using that technology expected to launch next year.