Klepsydra at Space Expo: Pushing the Boundaries of Safety-Critical AI for Space Across Multiple Hardware Accelerators
At Space Expo, Klepsydra is showcasing its latest advances in safety-critical AI, demonstrating how advanced AI algorithms and language models can be deployed efficiently across multiple hardware accelerator platforms for space applications.
Our work explores a broad range of technologies and approaches, including BrainChip’s neuromorphic technology, innovative new approaches to AI acceleration on FPGAs, and NVIDIA GPU platforms adapted to the demanding requirements of space systems. We are also demonstrating the execution of language models at the edge, opening new possibilities for intelligent onboard processing.
One of the highlights of our showcase will be a live demonstration of Klepsydra running on the Microchip High-Performance Spaceflight Computing (HPSC) processor, highlighting the potential of next-generation processing architectures for advanced onboard AI applications.
We will also present a realistic vision-based lunar landing application, running in parallel with an advanced AI-powered data compression algorithm. By combining autonomous visual navigation and intelligent data compression, the demonstration aims to enable live video streaming of the lunar landing back to Earth, overcoming the bandwidth constraints of space communications.
Through these developments, we are addressing the key challenges of next-generation space systems: high performance, low latency, energy efficiency and the reliability required by safety-critical applications.
From onboard data processing and AI-powered vision and navigation to language models and advanced hardware acceleration, discover how Klepsydra is pushing the boundaries of AI at the edge for space.
Save the date
When: November 17 – 19, 2026
Klepsydra Booth C55 in Hall 4
Our partners at Space Tech Expo
Agile FPGA Deployment
Klepsydra Technologies and ZHAW InES present a new, resource-efficient approach to FPGA-based AI acceleration, enabling dynamic DNN offloading from CPUs to FPGAs across Linux and RTEMS for space and edge applications.
Teledyne e2v LX2160
Klepsydra demonstrates parallel AI execution on the Teledyne e2v LX2160, combining real-time lunar hazard detection and advanced AI-based image compression for safer landings and live video streaming from the Moon to Earth.
Microchip HPSC
Klepsydra AI is successfully ported to Microchip’s HPSC/PIC64GX space-grade processors, leveraging hardware acceleration and multi-core execution to deliver high-performance, low-latency and secure AI for next-generation space applications.
Embedded Brains RTEMS QDP
Under ESA’s QSMART contract, Klepsydra is developing the first space-prequalified AI inference engine, leveraging Embedded Brains’ RTEMS 7.1 QDP. A live demo of Klepsydra AI on Gaisler NOEL-V for lunar landing will be showcased at Space Tech Expo.
BrainChip Demo
Klepsydra and BrainChip demonstrate autonomous space navigation powered by Akida neuromorphic processors and Klepsydra’s high-performance AI software. They also announce the first beta release of their integrated, ultra-low-power AI runtime for space applications.
LLMs in Space
Klepsydra and Thales Alenia Space Spain demonstrate onboard thermal sensor data analysis using LLMs, running on NVIDIA Jetson Orin and accelerated by Klepsydra’s GPU connector for next-generation Earth Observation missions.