Rama Afullo’s first reaction was, naturally: “Let’s go!” His company, Satlyt, is developing software designed to transform existing satellites into a data-processing network—essentially, an orbital data center. And he has just taken another step toward that goal.
Earlier this month, following 13 months of development, the company successfully completed a test. Conducted as part of a NASA-funded research project in collaboration with the University of Houston, the test demonstrated that Satlyt’s software could dynamically assign tasks to multiple computers in orbit, based on communication conditions and available computing resources.
The objective, Afullo explained, is to create software that can link multiple satellites to share computational workloads. He told TechCrunch that users simply need to “define permissions, priorities, and operational constraints,” and the software then makes decisions within those parameters. The software is already running on three satellites currently in orbit, and this test marks the first step toward realizing that broader vision. While the test utilized research funded by NASA, the agency has not yet selected the software for any specific mission or technology deployment.
Afullo noted that NASA missions face the very constraints Satlyt aims to address: “limited computing resources, intermittent communication, and the need to relay useful information back to people.”
For this test, the company created a simulated network comprising four computers, four users, and eight services. The software evaluated how to distribute and execute tasks across these computers, factoring in the availability of both computing and communication resources. A physical computer onboard a spacecraft ran the experiment—simulating a network of four computing hosts—and transmitted the results back to Earth. The test proved that Satlyt’s software can generate operational plans for orbital hardware; however, further testing is required to understand how information is transferred between multiple physical computers during actual operation.
“The results represent a series of trade-offs,” Afullo said, noting that one approach performed best but also demanded more computing power than the others.
Nevertheless, for Satlyt—which placed in the top five at this year’s TechCrunch Disrupt Battlefield competition—the outcome is a clear victory. “This is the cornerstone for satellite-based shared computing tasks,” the company head stated.
NASA already possesses a technical network capable of transmitting data in space, even when communication links experience delays. NASA began utilizing this technology earlier this year and has already conducted tests aboard aircraft.
Afulo, who previously worked at Google and SpaceX, founded the company Satlyt in 2024 with the goal of transforming spacecraft into virtual data centers. The company’s software enables operators to run artificial intelligence and other applications directly on spacecraft, allowing for data processing closer to the point of collection. This reduces reliance on downlinks—the transmission of data from satellites back to Earth—and subsequent ground-based analysis.
For instance, if a spacecraft detects a wildfire on Earth, such software can help it relay relevant information more rapidly. “Our objective is to deliver useful information to people faster and to make better use of limited communication windows,” he said. “This experiment tests one planned component of that broader capability.”
Satlyt stated that they still need to test their software across various workloads, hardware configurations, and connectivity conditions. Afullo also noted an issue the team discovered regarding how the software handles latency: it records latency as an “arrival timestamp” (the actual time of arrival) rather than as “elapsed latency” (the duration taken for data to arrive)—two distinct metrics.
However, this does not mean the experiment was a failure. “The software development, on-board execution, and output results were all successful,” he said.
Afullo added that there is still much work to be done. “We still require functional communication links, authorization from participating operators, secure transmission, resource scheduling, and recovery mechanisms for instances where the computer connection is interrupted.” Satlyt is also exploring applications beyond low Earth orbit—including a NASA-funded project aimed at supporting the transmission of 4K video from the Moon.
“Our next technical steps involve broader validation and integration, followed by a demonstration of sharing computational tasks between different spacecraft,” he said. “In the long term, we want developers to be able to build space-based applications without having to rebuild the infrastructure for every single mission. That is the vision behind what we call a ‘virtual AI data center’ in space.”