AI Sending Messages From Space

For decades, satellites have circled Earth collecting images, but much of the data never reaches the ground or arrives too late to be useful. Zaitra, a Czech startup based in Brno, is trying to change that. Its Skaisen AI model processes data in orbit and beams down concise, machine-readable alerts, cutting latency from hours or days to minutes.

VP of Engineering Martin Javorka(Left), CEO Marek Marusin(Middle), CTO Rado Pitonak(Right)

Zaitra

Co-founder and CEO Marek Marusin calls this “selling insights, not images.” Traditional Earth-observation companies compete to deliver sharper or more frequent pictures. “We don’t send the picture,” Marusin says. “We send what the AI sees in the picture.” The result is a near-real-time alert that might read: “there are three untracked vessels at these coordinates, heading north at this speed.” That alert can cue follow-up satellites, trigger drone inspections, or simply send a message to an operator’s phone.

CTO Rado Pitonak explains the company builds its own computers based on FPGA (Field Programmable Gate Array) architecture. “Every ounce is critical,” he says. “We had to build hardware that is light enough, power-efficient, and space-qualified, then run large models on it.” The result is Skaidock, a ruggedized processor that hosts Zaitra’s Skaisen software. Together, they filter out unusable images, discard cloud-covered scenes, and transmit only structured text that other systems can read and act on.

Zaitra demonstrated its first prototype in space in July 2022 and has since integrated Skaisen on two commercial missions. The company has sold about a dozen Skaidock units this year.

Zaitra’s algorithm was tested on the Czech VZLUSAT-2 satellite, where it filtered out cloud-covered images and transmitted only useful data, proving its ability to save bandwidth and deliver cleaner information to users on the ground. Its hardware and software also flew aboard Troll, the largest Czech satellite yet, where they successfully faced full operational trials. Zaitra recently presented its edge-processing approach at the SmallSat Expo in Utah, one of the industry’s leading gatherings.

Use cases go far beyond defense. Zaitra is working on methane-leak detection for pipeline operators, wildfire ignition point identification for emergency responders, and illegal fishing enforcement. “Once you have near-real-time information, you can automate things,” Marusin says. “You can launch a drone, redirect a ship, or just show the data on a commander’s map. That was not possible before we started using AI on top of satellites.”

Like many space startups, Zaitra is pursuing a two-track strategy: selling its components to other satellite makers while building its own constellation. The company closed a €1.7 million pre-seed round last year and is now opening a seed round to fund its first Zaitra-operated AI node and a U.S. expansion. The team has grown from five to 25 in the past year and has signed roughly ten paying customers.

The story of how Zaitra began could have come from a movie. “We went to a hackathon for fun, food, and curiosity,” Marusin recalls. “We won, and the organizer liked the team so much he invested the first money into the company.” They pivoted from agency work to building their own AI hardware and software stack, and Marusin landed on Forbes’ 30 Under 30 list. The advisor is Petr Kapoun, who founded satellite system integrator company TRL Space. He is a founding angel investor in Zaitra.

Marusin says Zaitra’s addressable market for components is roughly $1 billion, but the market for low-latency space data is far larger, and growing. “The total market for space data is about $5.5 billion today,” Marusin says. “It can grow into the trillions as we find new applications.”

“Five years ago, we didn’t have the hardware or the AI to do this,” he says. “Now we can send a message from space in minutes.”