Payload programmes
Extract more mission value from every observation by filtering, classifying and prioritising data before downlink.
AIQSAT project · Technology
Onboard AI Computing
Radiation-tolerant edge processing for autonomous spacecraft
We are developing this technology as part of an intelligent, quantum-secure satellite network for stable data exchange between orbit and Earth.
Onboard AI computing
What we are developing
An intelligent satellite needs to decide in orbit, not after the next downlink. AIQSAT builds the onboard computing stack: radiation-tolerant AI accelerators, model compression and a flight software layer that lets neural networks run safely next to the platform's critical functions.
Radiation-tolerant accelerators
Space-qualified FPGA and SoC based inference hardware with error correction, watchdogs and redundant boot paths for the harsh orbital radiation environment.
Compressed neural networks
Quantised and pruned vision and time-series models that fit into a few watts of power budget while keeping detection accuracy close to ground-grade models.
Autonomy and edge triage
Onboard cloud masking, target detection and anomaly flagging, so only relevant scenes and events consume precious downlink bandwidth.
Secure model updates
Signed over-the-air model uploads, staged rollout and in-orbit A/B verification before a new model is given operational authority.
Technology stack
- Rad-tolerant FPGA / SoC inference
- TinyML & quantised CNN / transformer models
- Federated and incremental in-orbit learning
- Fault-tolerant real-time flight software
The opportunity
Move intelligence and trust closer to the source
A satellite that must send every raw observation to Earth is limited by contact windows, bandwidth and ground response time. As sensors become more capable and constellations grow, centralised processing becomes an operational and economic bottleneck.
AIQSAT is developing a dependable computing layer that lets spacecraft interpret conditions, select valuable data and support mission decisions in orbit—while keeping safety-critical control isolated, observable and under human authority.
System architecture
A connected path from orbital signal to trusted decision
The project is designed as a modular system. Each layer can create value independently, while the complete architecture enables a secure, intelligent data network across orbit and Earth.
- 1Ingest
Unify payload and health data
A controlled interface accepts imagery, signals and telemetry without exposing critical flight functions.
- 2Infer
Run efficient models
Quantised models execute within strict power, memory, thermal and timing budgets.
- 3Assure
Check every decision
Confidence thresholds, watchdogs and deterministic fallback paths limit operational risk.
- 4Improve
Update with control
Signed packages and staged activation allow models to evolve over a mission's lifetime.
Strategic value
Built around the needs of future operators
AIQSAT is being shaped for organisations that need resilient infrastructure, faster access to useful information and a security model designed for the quantum era.
Fleet operators
Move toward event-driven tasking and scalable supervision without removing operational safeguards.
Platform manufacturers
Offer customers a differentiated, reusable compute capability alongside the spacecraft bus.
Strategic investors
Back a software and systems layer with applicability across observation, communications and security missions.
Development path
A staged route from validation to deployment
The roadmap is structured to retire technical risk early, prove interoperability and create clear decision points for partners and capital providers.
Bench
Workload characterisation
Benchmark priority models against representative data, hardware limits and mission assurance criteria.
Flight-like
Resilient compute module
Integrate accelerator, fault detection, secure boot and thermal management in a representative environment.
Orbit
Controlled demonstration
Validate inference quality, energy use, radiation response and update procedures in flight.
Product
Reusable mission layer
Package qualified workflows and interfaces for adoption across spacecraft and payload families.
Investment case
An infrastructure opportunity at the intersection of three markets
The project combines differentiated engineering, reusable intellectual property and a platform approach intended to support multiple missions, customers and commercial models.
Cross-mission IP
The value lies in optimisation, assurance and orchestration that can be reused beyond a single hardware configuration.
High integration value
A validated onboard stack can reduce customer effort at the difficult boundary between models, compute and flight operations.
Recurring lifecycle role
Potential revenue extends from integration to licensed runtime, controlled model updates and mission support.
Build the next layer with us
A more efficient, resilient and secure future for orbital data.
We welcome compute suppliers, mission owners, AI specialists and investors to help move trustworthy orbital intelligence from laboratory validation into flight.
More technology
