In April, a satellite built by Loft Orbital made history by identifying objects without human intervention. The spacecraft, Yam-9, used a vision-language model from Google DeepMind to locate areas of interest in response to natural language queries. This marks the first reported use of a vision-language model in orbit, according to TechCrunch. Typically, satellites transmit large data sets to analysts on Earth, who then use machine learning or manual methods to interpret the data. Yam-9’s onboard software, developed by NASA’s Jet Propulsion Laboratory, allowed it to analyze imagery and respond to queries about specific environments or infrastructure. The demonstration is significant for two reasons. In the near term, it could make space sensors more useful by performing initial data triage on orbit, reducing the amount of raw data analysts must process. In the long term, it represents a step toward running larger-scale AI infrastructure in space. Loft’s head of AI, Paul Lasserre, said the technology opens the door to always-on, patrol layers in space, allowing satellites to monitor borders or detect suspicious activity. Loft’s spacecraft are designed as platforms for third-party customers, operating under an infrastructure-as-a-service model. One recent deal involved building, launching, and operating six new satellites for EarthDaily, which will analyze and market the data collected onboard. Yam-9 was launched in the fall of 2025 as a pathfinder for Loft’s orbital AI projects, and includes a Nvidia Jetson Orrin AGX GPU, one of the leading chips used in space compute. Juan Delfa Victoria, a technical leader in NASA JPL’s AI group, led the development of NAVI-Orbital, a software package that served as the harness for the Gemma 3 VLM. While Gemma 3 is off the shelf, software engineers had to streamline the package to reduce the amount of libraries and memory it would require. The goal is to build out the constellation to ensure real-time coverage of anywhere on Earth, which would take between 50 and 100 satellites like Yam-9. Loft currently operates 12 spacecraft on orbit. Lessons learned from deploying these smaller models on orbit will inform how companies attempt to deploy larger-scale compute infrastructure in space, particularly in the areas of power and memory management. They could also pave the way for new scientific tools. The idea for NAVI-Space began with JPL Researcher Taran Cyriac John, who was thinking about digital assistants for astronauts exploring the Moon or Mars. “We’re thinking, okay, you have astronauts with pressurized suits, and you know they cannot be tapping on a keyboard, whatever they want to do is complex,” Delfa Victoria said. “So, how about we provide an assistant, like in video games and in movies, where you see an AI which is interactive?” Just don’t call it HAL 9000.

Source: techcrunch