Klepsydra AI Redefines Satellite Autonomy: Successful Completion of Final Review in European Space Agency Initiative

 

Zug,Switzerland — January 12, 2026 — Klepsydra today announced the successful completion of the Final Review for the PATTERN project, marking a major leap forward in bringing high-performance Artificial Intelligence to space-qualified hardware. Supported by Frontgrade Gaisler under a European Space Agency (ESA) initiative, Klepsydra has successfully demonstrated an AI inference engine that fulfills ECSS standards to criticality level ‘D’ and performs up to 8x faster than existing industry standards.

As satellite missions grow in complexity, the need for on-board real-time decision-making is critical. The PATTERN project addresses this by enabling satellites to process data locally, reducing reliance on ground stations and accelerating insights for Earth observation, autonomous navigation, and collision avoidance.

Breaking Hardware Barriers

The final phase of the project has successfully extended Klepsydra AI’s compatibility across the cornerstone of European and international space processors. The software is now fully ported and optimized for:

• Gaisler LEON4, LEON5, and NOEL-V (RISC-V)
• Microchip Polarfire
• RTEMS6 SMP and VxWorks Operating Systems

“By successfully porting Klepsydra to all five different space-grade processor architectures, reaching ECSS standards to criticality level ‘D’, and achieving exceptional performance results on multi-core architectures, we are providing the space sector with a solution that is not only faster but significantly simpler to deploy,” said Klaus Buchheim, Head of Business Development at Klepsydra.

Unprecedented Performance Gains

The PATTERN project results showcase a dramatic improvement over traditional frameworks like TensorFlow Lite Micro (TFLM). Key performance benchmarks include:

• 8x Speed Increase: On 4x core configurations compared to TFLM.
• Optimization Mastery: Up to 6x faster execution on NOEL-V using specialized SIMD mathematical backends.
• Streamlined Deployment: A new workflow allows developers to convert models to the “KPSR” binary format, eliminating heavy external dependencies and drastically reducing “time-to-flight.”

The Path Ahead

With the Final Review accepted by project leadership, Klepsydra is moving immediately into the next phase of development. The immediate next step is to move into criticality level ‘C’, including unique features like FDIR (Fault Detection, Isolation, and Recovery) for AI applications—a capability not available in any other inference engine currently on the market. These enhancements will ensure the highest level of mission-critical reliability for autonomous space operations.

About Klepsydra Klepsydra is a leading provider of high-performance software for edge computing. Inspired by high-frequency trading techniques, Klepsydra’s technology optimizes data processing for space, robotics, and automotive industries, enabling real-time AI in the most resource-constrained environments.

Media Contact: isabel.delcastillo@klepsydra.com