AI systems
Data quality management
Measuring and managing completeness, accuracy, representativeness and timeliness of training and input data.
We design and operate.
ISO/IEC 5259 series
Benchmarks: models and radios measured on the same accelerated rack under the same load. The loop, step by step: 1. A rack, a GPU server and a results screen. 2. A standard workload is loaded. 3. The server runs it at full load. 4. Time and place are fixed to every sample. 5. Timings stream to the benchmark log. 6. Relays carry the packets without reading them. 7. Signatures are checked before any model runs. 8. Throughput and power are normalised. 9. The edge releases confirmed results only. 10. The leaderboard updates on screen. 11. The console shows the evidence with the alert. 12. The run is repeated for a stable median. 13. Sensors check that the response worked. 14. The whole chain is archived for audit. 15. Everything settles back to idle.
01 / 15
A rack, a GPU server and a results screen
Nexus Observatory/
Frontier labs/
BenchmarksOpen data, one method, comparable results.
Benchmarks, inside Frontier labs: data quality management, experiment tracking and reproducibility, model and dataset bills of materials, robustness testing and more. Every item is traced to its specification, from ISO/IEC 5259 series to ISO/IEC 23053, and runs on the open Observatory Protocol.
Plan with usArchitectureThe protocol- Technologies
- 7
- At the frontier
- 0
- Protocol functions
- 3 of 12
- Industries served
- 3
01What it is
MLOps.
AI from model to accountable operation: edge inference and model optimisation, foundation models for Earth observation and geospatial work, time-series and sensor AI, multimodal fusion and supervised agentic operations, federated and privacy-preserving learning, physical AI and robotics within stated red lines, MLOps, evaluation and model risk, and the governance frameworks that apply to each. Each technology below carries what it is, what we do with it, and the standard it is drawn against.
AI systems
Experiment tracking and reproducibility
Recording code, data, parameters and environment so any model can be rebuilt.
We design and operate.
ISO/IEC 23053
AI systems
Model and dataset bills of materials
Inventories of models, datasets and dependencies for each deployed AI system.
We design and operate.
ECMA-424 (CycloneDX, ML-BOM) · SPDX 3.0 (AI and dataset profiles)
AI systems
Robustness testing
Testing under noise, shift, missing inputs and edge cases.
We design and operate.
ISO/IEC TR 24029-1 · ISO/IEC 24029-2 (robustness of neural networks)
AI systems
Bias assessment
Measuring unwanted differences in performance across groups and regions.
We design and operate.
ISO/IEC TR 24027 (bias in AI systems)
AI systems
Adversarial testing and red-teaming
Probing models for evasion, poisoning, prompt injection and data leakage.
We design and operate.
NIST AI 100-2 (adversarial machine learning taxonomy) · OWASP Top 10 for LLM Applications
AI systems
Training data and labels
Documented, licensed training sets with known label quality and geographic coverage.
We design and operate.
OGC TrainingDML-AI · ISO 19157-1 (data quality)
02Architecture
Where it sits, and what it speaks.
Benchmarks draws on the spectrum below; the highlighted layers carry this area, and the protocol functions it speaks are marked.
Edge inference, model serving and optimisation
10
Foundation models for Earth observation and geospatial
9
Time-series and sensor AI
8
Multimodal fusion and agentic operations
7
Federated and privacy-preserving learning
7
Physical AI and robotics (within red lines)
7
MLOps
7
Evaluation, model risk and AI governance
10
O-RAN architecture and open interfaces
7
Fronthaul (7.2x split) and transport
8
RAN intelligent controllers, xApps and rApps
8
SMO, O-Cloud and accelerated L1
8
AI-RAN: shared accelerated infrastructure
7
AI/ML in the air interface and network management
8
Energy saving and efficiency
6
Testing, integration and security
8
01Join Identity, attestation quote, signed policy.
02Admit Admitted for a stated scope, nothing wider.
03Describe What, where, how and by whom, with calibration.
04Time One clock across the grid, with its error bound.
05Observe Every value with quality and provenance.
06Qualify Tested against the specification, state on record.
07Federate Signed exchange across grids and trust domains.
08Orchestrate Software and configuration by signed plan.
09Correct A correction marks what depends on it and publishes a new version.
10Exchange Products reach people and applications through versioned interfaces.
11Prove Signed proof receipts that an independent verifier can check.
12Exit and decommission An open exit path, and a record that closes.
03On the grid
Every site joins one observing grid.
The open networks on the grid today, over real relief and the geodesic cells we count them into, filtered to the space domain this area serves. A site built here enrols into the same grid.
The observing grid
- Open space instruments on the observing grid.
04In the field
Where this runs, sector by sector.
Deployment scenarios from the sector research that rely on benchmarks.
ScientificResearch-infrastructure gap-filling nodes
Observatory-as-a-service for research infrastructures that must extend coverage under their own quality rules.TransportNon-federal automated weather observing for small airports
Aviation-grade stations under the civil-aviation authority's advisory circular; pricing not researched.Defence-adjacent (civil dual-use)Research and development route through the defence science programme
A science-and-technology programme on all-domain situational awareness.05Open and enterprise
Open by default. Enterprise by contract.
The open counterpart lives on the main site for anyone to read and run. The enterprise service is designed, deployed and operated with you, under contract.
Open, on the main site
- Benchmark methods
- Pilot reports, published either way
Enterprise, on this site
- Research pilots
- Prototype engineering
06The rest of the line
Frontier labs, area by area.
Frontier labsPilots
New technology, a real site, a written plan.Frontier labsAI-RAN and 6G
Sensing and communication on shared compute.Frontier labsSpace and NTN
Direct-to-device and new payloads, tested.Frontier labsPhysical AI
Drones and robots, inside published red lines.Frontier labsInto the product
From a result to a qualified release.Bring us the site.
Survey, design, deployment and operation for benchmarks. The form names who receives your details.
Talk to usAsk for a quoteMore presentations
See it from every side.
Each opens as a full-screen presentation: deeper pages on the same subject, explainers, stories, guided tours of the places, network scenes and the observatories that came before.
All presentations for this page