Satlyt Achieves Key Milestone in Quest to Turn Orbiting Satellites Into Distributed Data Centers

When founder and CEO Rama Afullo first received the telemetry data confirming that his startup’s latest orbital software test had succeeded, his reaction was instantaneous and enthusiastic. "Let’s GOOOO," he said.

For Afullo and his team at Satlyt, the successful test represents a major, albeit preliminary, step toward realizing an ambitious vision: transforming existing satellites in low Earth orbit into a cooperative, interconnected data processing network. By turning ordinary spacecraft into virtual orbital data centers, the company hopes to fundamentally alter how space-based assets handle, analyze, and transmit massive streams of data back to Earth.

The successful trial, completed earlier this month after 13 months of rigorous development, was conducted in partnership with the University of Houston as part of a NASA-funded research project. Rather than a full-scale multi-satellite deployment, the test was designed to evaluate how Satlyt’s proprietary software could intelligently distribute computing tasks among multiple virtual nodes in orbit based on real-time communication availability and available computing resources.

Afullo, a former product manager at both Google and SpaceX, launched Satlyt in 2024 to address what he views as one of the most stubborn bottlenecks in modern space exploration and commercial satellite operations: the heavy reliance on ground-based data analysis and limited downlink windows.

Currently, satellites capture vast amounts of data—ranging from high-resolution imagery and scientific telemetry to environmental monitoring—and must transmit those raw bytes back to ground stations for processing. This traditional approach is frequently hindered by intermittent communication links, limited bandwidth, and narrow geographic windows where a satellite has a direct line of sight to a ground receiver.

Satlyt’s software is designed to bypass these limitations by enabling spacecraft to run artificial intelligence and other intensive applications directly onboard. By processing information closer to the point of collection, a satellite can immediately identify critical events—such as spotting a sudden wildfire or tracking an unfolding natural disaster—and transmit actionable intelligence rather than raw, unprocessed data.

"The goal is to get useful answers to people sooner and make better use of limited communication windows," Afullo explained. "This experiment tested one planning component of that broader capability."

During the recent 13-month development cycle, the Satlyt team built and executed a simulated space network for the test. The experiment modeled an environment consisting of four computing hosts, four distinct users, and eight unique services. The startup’s software was then tasked with assessing the optimal way to assign and run various tasks across those simulated computers, continuously accounting for shifting compute loads and communication availability.

To execute the test in the harsh environment of space, a single physical computer aboard an active spacecraft ran the experiment, evaluating the simulated four-node network before downlinking the results back to Earth. Satlyt’s software is already installed on three operational satellites currently in orbit, laying the groundwork for more advanced demonstrations in the future.

The primary takeaway from the recent test, according to Afullo, was a nuanced engineering trade-off. The software successfully generated operational plans on hardware in orbit and managed the workload, but the evaluation revealed that the most effective approach required significantly more computing power than alternative methods.

Even with these complexities to iron out, Satlyt—which earned a spot in the top five of TechCrunch Disrupt Battlefield this year—considers the trial a resounding victory. "It’s a building block toward satellites sharing computing work," Afullo noted.

While the research project was supported by NASA funding, the space agency has not yet selected Satlyt for any specific mission or chosen the startup’s technology for immediate operational deployment. However, Afullo emphasizes that NASA missions face the exact constraints that Satlyt’s technology is engineered to solve: limited computing resources, intermittent communication windows, and an urgent mandate to deliver actionable information to human operators as quickly as possible.

Furthermore, the startup’s NASA-funded research extends beyond low Earth orbit. Satlyt is also exploring applications in deep space, including potential support for high-definition 4K video streaming transmissions from the Moon.

Despite the success of the recent test, the team acknowledges that a substantial amount of development and validation remains ahead. During the evaluation, the startup identified a subtle data-tracking nuance regarding how the software time-referenced communication delays. The system was inadvertently stamping delays based on arrival time rather than elapsed delay, representing two distinct operational metrics.

Afullo was quick to clarify that this technical discrepancy did not constitute a failure of the experiment. "The software development, onboard execution, and return of outputs succeeded," he stated, highlighting that identifying such edge cases is precisely the purpose of rigorous in-orbit testing.

Moving forward, Satlyt plans to subject its software to a wider variety of workloads, diverse hardware configurations, and fluctuating connectivity environments. Afullo points out that true distributed computing in space requires more than just smart task-allocation algorithms; it demands reliable communication links, strict permissions protocols managed by participating satellite operators, secure data transfers, dynamic resource scheduling, and robust recovery mechanisms for instances when a computer node drops offline.

The startup’s immediate roadmap calls for broader technical validation and system integration, culminating eventually in live demonstrations where computing tasks are shared simultaneously across physically separate spacecraft operated by different entities.

"Our next technical steps are broader validation and integration, followed by demonstrations of computing work shared across separate spacecraft," Afullo said. "Longer term, we want developers to build applications for space without having to rebuild the infrastructure for every mission. That’s what virtual AI data centers in space mean to us."