Rama Afullo expressed his enthusiasm with a spirited “Let’s GOOOO” as his company, Satlyt, advances its mission to transform existing satellites into data processing networks akin to orbital data centers. This month marked a significant milestone toward that goal.
After a year of dedicated effort, Satlyt successfully conducted a test earlier in the month. Collaborating with the University of Houston on a NASA-funded project, the test demonstrated the capability of the startup’s software to efficiently distribute tasks among orbiting computers based on available communication and computational resources.
According to Afullo, the objective is to develop software that interconnects various satellites to streamline computing tasks. Users simply need to specify permissions, priorities, and operational boundaries, allowing the software to autonomously manage tasks within those parameters. This software is already operational on three satellites in space, and this recent test represents a key step toward achieving that vision. Although the software was tested using NASA’s funding, the agency has not yet identified it for specific missions or deployments.
Afullo noted that NASA’s missions encounter similar challenges that Satlyt aims to solve, such as limited computational capacity, sporadic communication, and the necessity of delivering actionable information to users.
In the test, Satlyt simulated a network with four computers, four users, and eight services. The software evaluated how to allocate and execute tasks among these computers, considering both computational and communication availability. Only one physical computer aboard a satellite executed the software, while the others were simulations, with outcomes relayed back to Earth. The test confirmed Satlyt’s ability to create operational plans for in-orbit hardware, but further evaluations are necessary to explore how information will be exchanged between multiple physical computers in practice.
Afullo described the findings as presenting a trade-off, revealing that one method performed best, albeit requiring more computational resources than others.
Nevertheless, Satlyt, a frontrunner in the TechCrunch Disrupt Battle, views this as a significant achievement. “It’s a fundamental step toward enabling satellites to collaboratively handle computing tasks,” he stated.
NASA has already implemented a technology network capable of transmitting data in space even when faced with communication lags, having begun tests on aircraft earlier this year.
Afullo, who has a background at Google and SpaceX, established Satlyt in 2024 to enable spacecraft to function as virtual data centers. The software assists operators in executing AI and other applications onboard, allowing data processing to occur nearer to its source and reducing reliance on terrestrial data retrieval and human-based analysis.
For instance, if a satellite detects a wildfire, this software can expedite the communication of critical information. “Our aim is to deliver actionable insights more swiftly and optimize limited communication opportunities,” Afullo remarked. “This experiment focused on one aspect of that larger functionality.”
Satlyt remains committed to testing its software in various operational conditions, including different workloads, hardware setups, and connectivity scenarios. Afullo mentioned they encountered a specific issue related to their software’s handling of time delays; it was incorrectly labeling the arrival time as the total delay, which misrepresented the data.
Nonetheless, this does not indicate a failure of the experiment, as Afullo asserted. “The development of the software, its deployment, and the retrieval of results were all successful,” he stated.
However, Afullo acknowledged that significant work is still required. “We also need reliable communication channels, necessary permissions from operators, secure data transfers, resource management, and recovery processes for instances when computer connections are lost.” They aspire to expand their operations beyond low Earth orbit in the near future.
“Our forthcoming technical actions involve broader validation and integration, followed by demonstrations of distributed computing across separate spacecraft,” he elaborated. “In the long term, we envision enabling developers to create applications for space without the need to reconstruct the infrastructure for each mission. This vision of virtual AI data centers in space is what we strive for.”
