Skip to main content

Meridian Space Documentation

Meridian Space builds high-performance, radiation-tolerant Edge AI infrastructure for the next generation of autonomous satellite constellations.

This documentation provides data-driven benchmarks of our hardware and software stack, focused on real-world space applications: Earth Observation, Vision-Based Navigation, and onboard inference at the edge.

What We Build​

  • Odin v0 — Edge AI compute module for SmallSats. Jetson Orin Nano Super paired with the Axelera Metis M.2 D-IMC accelerator. 15 TOPS/W efficiency, 10–20 W system power envelope, 1U form factor.
  • HALO (Hardware Abstraction Layer for Orbit) — Software bridge between PyTorch models and Odin hardware. GPU and IMC-AIPU run in parallel.
  • SNR (Space Native Runtime) — Software-level radiation mitigation with optimized inference scheduling.

Benchmarks​

Real benchmark data across three satellite workloads — GEMM throughput, spacecraft pose estimation, and multi-spectral cloud detection. All comparisons are head-to-head against NVIDIA Jetson TensorRT on identical hardware.

→ View Benchmarks