Skip to content

SO3 Innovation, Royal Naval Reserve / Current / Reserve role, active SC clearance

Defence innovation

An AI scenario generation tool in use at divisional level, and a sovereign air-gapped LLM platform, designed and specified.

  • AI scenario generation
  • Local LLMs
  • Retrieval
  • Audit logging

The concept in two objects. Consumer Apple silicon and open-weight models such as Gemma, running entirely offline. Illustrative, since the platform is designed and specified, not built.

01 · Problem

Defence training and knowledge work have the same AI needs as industry, with constraints industry never sees. Sovereignty. Audit. Networks that cannot touch the internet.

02 · What I built

  • An AI scenario generation tool for designing training serials. Built in my innovation role, in use at divisional level.
  • A sovereign, fully air-gapped LLM platform, designed and specified. Local open-weight models such as Gemma on Apple silicon hardware, retrieval over reference material, full audit logging, no external APIs.

03 · Architecture

The air-gapped platform as designed and specified. It has not been deployed. Nothing crosses the boundary except vetted physical media.

04 · What happened

The scenario generator is in use at divisional level. The air-gapped platform is designed and specified, awaiting go-ahead.

Work carried out in my reserve role.