Edge AI for satellites, rovers & HAPS
Inference on-orbit. No downlink, no waiting, no compromise.
Single-event-upset tolerance, redundant compute fabric, and a thermal envelope qualified for sustained vacuum. Run frontier models on the platform — for satellites, lunar rovers, and high-altitude systems where the downlink is the bottleneck.
Use cases
Where on-platform AI earns its keep
On-orbit inference
Triage Earth-observation frames before downlink. Send the conclusions, not the raw pixels — and free the bandwidth budget for what actually matters.
Onboard scene understanding
Vision-language reasoning on rovers and orbital platforms — caption a scene, identify anomalies, and act without a ground-loop in the way.
Lunar & planetary autonomy
Long-duration autonomy across communication windows that can stretch from seconds to hours. The model survives the gap.
Anomaly & fault detection
Run live anomaly detectors on raw telemetry. Catch failures inside the platform before the next ground pass.
Radiation-tolerant compute
TMR + ECC + scrub for SEU mitigation. 20 krad TID. Sustained operation through the regimes that retire commercial silicon.
Vacuum-qualified thermal envelope
Operate −55 to +105 °C, MIL-STD-883 thermal cycling, 10⁻⁶ Torr vacuum-qualified packaging.
The matched module
AstroCore S
FPGA-based · space-grade
AstroCore S brings on-board AI to satellites, lunar rovers, and high-altitude platforms. Built on the Invotet Unified Engine in a radiation-tolerant FPGA fabric — single-event-upset tolerance, redundant compute, and a thermal envelope qualified for sustained vacuum operation. Deploy-ready hardware, not a chip-down ASIC waiting on tape-out.
- Throughput
- 38 GOPS
- Power envelope
- 4.5 W
- Operating range
- −55 to +105 °C
- Tolerance
- 20 krad TID
Why this module, this vertical
The properties that map to your platform.
Sustainable autonomy
Frontier-class models inside a sub-15W envelope. AI fits inside the battery or solar budget — Size, Weight, and Power optimized for every module.
Transformer-grade fidelity
BF16-native execution preserves training-equivalent accuracy at 95% sustained utilization, with native flash attention and hardware tensor-parallel sync. A cycle-accurate hardware trace buffer and compile-graph-to-hardware specialization make every inference verifiable and tuned to the workload.
Mil-spec environment
Operate from −55 °C to +105 °C. Survive thermal cycling, vacuum, and radiation regimes that disqualify commercial silicon.
Secure by design
Hardware root of trust, signed firmware, and per-module attestation keep both the model and the device tamper-evident.
Invotet SDK
Compile once. Deploy to every Invotet module.
A unified Python SDK that ingests PyTorch, ONNX, and HuggingFace checkpoints, quantizes for Invotet modules, and ships a deterministic runtime to the device. No CUDA in the loop.
Framework
PyTorch
Trace or torch.export checkpoints compile directly with no rewrite.
Framework
ONNX
Standards-based interchange — compile any ONNX-exported model.
Framework
HuggingFace
transformers checkpoints land on Invotet through a one-line loader.
Talk to the team that ships the modules.
Most space conversations start with a qualification report and a thermal/radiation profile. Tell us your bus, the orbit, and the mission window — we will route the right datasheet and qual evidence the same day.