Intended roles
- Platform engineers
- DevOps engineers
- Security engineers
- Enterprise architects
GatedOutlineadvanced program
Air-Gapped AI Platform Engineering
Build the controlled supply chain and runtime required to operate AI workloads without direct runtime internet access.
- Version
- 0.1.0
- Modules
- 7
- Delivery
- Pilot planning
- Availability
- Curriculum and lab under development
- Content review
- 2026-08-02
- Automated lab
- Not yet validated
Program fit
Built for teams responsible for implementation decisions.
A pilot program for building the controlled artifact supply chain and runtime required to operate AI without direct internet access.
Prerequisites
- Hands-on container and Linux administration experience
- Working knowledge of Kubernetes or OpenShift
- Understanding of enterprise PKI, registries, and software supply-chain controls
Capabilities
What the program is designed to develop.
Design a connected import and approval zone
Operate internal registries and package sources
Validate and promote model artifacts
Build and update workloads offline
Produce supply-chain and runtime evidence
Shared module sequence
One maintained registry, assembled for this outcome.
Program pages resolve module titles and descriptions directly from the Academy registry. Updates remain controlled in one source rather than copied across pages.
- 1SAI-100Content reviewed
Sovereign AI Foundations
Establish the vocabulary, deployment boundaries, shared responsibilities, and control objectives needed to reason about sovereign AI systems.
Open documentation - 2SAI-110Content reviewed
Trust Boundaries and Threat Modelling
Model data flows, actors, assets, attack surfaces, trust zones, and risk scenarios for private AI workloads.
Open documentation - 3SAI-200Content reviewed
Internal Artifact Supply Chains and Offline Builds
Control the import, verification, approval, storage, promotion, and offline build of AI software and model artifacts.
Open documentation - 4SAI-210Draft
Model Selection and Lifecycle Management
Evaluate model fit, licensing, provenance, packaging, approval, updates, and retirement inside a controlled lifecycle.
Open documentation - 5SAI-220Content reviewed
Model Serving, Routing, and Hardware
Select and operate model runtimes, routing patterns, hardware profiles, capacity controls, and reliability targets.
Open documentation - 6SAI-250Prototype
Governance, Assurance, and Evidence
Turn policies and control objectives into verifiable runtime, release, decision, and audit evidence.
Open documentation - 7SAI-270Draft
Kubernetes and OpenShift Operations
Operate restricted AI workloads with controlled networking, storage, GPU access, security contexts, updates, backup, and recovery.
Open documentation
Practical labs
Observable work inside an approved environment.
Exercises use local, customer-hosted, or explicitly approved private infrastructure and are designed to produce repeatable validation and evidence.
LAB-AIRGAP-01Outline
Disconnected Platform Pilot
Promote approved artifacts into a simulated disconnected runtime and perform a controlled update and recovery.
Lab outputs
- Internal artifact hub
- Offline runtime
- Update and recovery evidence
Last content review: 2026-08-03Last automated lab validation: Not yet validatedIndependent reproduction: Not yet completed
Lab boundary principles
- No customer documents or prompts are sent to SovAIHub by default
- Prerequisites and supported configurations are explicit
- Validation is automated or clearly repeatable
- The environment can be reset and the exercise repeated
- Limitations and unsupported configurations are stated honestly
Capstone and assessment
Finish with implementation artifacts, not attendance alone.
Capstone outputs
- Connected import-zone design
- Security and approval gate
- Internal artifact hub
- Container registry and Python wheelhouse
- Model import and validation record
- Offline build factory
- Air-gapped runtime
- Controlled update process
- Audit and evidence output
Assessment method
Pilot completion requires a repeatable disconnected deployment and evidence review. Availability depends on independent reproducibility testing.
Delivery modes
- Pilot planning
- Customer-hosted isolated-lab planning
- Platform readiness workshop
Credential guardrail: Initial delivery may use “program completed,” “assessed completion,” or “capstone passed.” It does not award a professional certification.
Connected implementation assets
Build on SovAIHub kits and solution patterns.
Program pages resolve kit details from the existing SovAIHub product registry, keeping product information maintained in one place.
Open Source Kit
SovAI Air-Gap AI Starter
Open-source laptop-ready starter kit for demonstrating an air-gap-ready sovereign AI runtime with local documents, approved tools, offline Docker bootstrap, and audit logs.
View kitLocal LLM RAG Kit
SovAI Air-Gap AI Starter v0.2 with Ollama
Open-source local LLM RAG starter that runs private retrieval, grounded prompting, Ollama inference, citations, and audit logging without an external LLM API.
View kitArtifact Hub
SovAI Air-Gap Internal Artifact Hub
Open-source reference implementation for a controlled internal AI artifact supply chain with local registry, wheelhouse, prompt/tool manifests, approvals, and offline builds.
View kitCurrent limitations
What this program does not claim yet.
Honest release boundary
- This is a pilot, not a generally available program.
- The lab has not yet passed the required independent reproducibility test.
- Supported cluster, runtime, package, and model versions will be narrow and explicitly declared.
- No production environment should be changed as part of an Academy pilot.
Enterprise pilot planning
Discuss Air-Gapped AI Platform Engineering for your team.
Share role mix, environment, technical constraints, data boundary, and desired implementation outputs. Do not submit sensitive architecture details through the public form.
No public checkout, account, or sensitive architecture upload required