Live Training Package specifications
The source of truth for compliance alignment. Every audit and rewrite is mapped against the current published unit specification on training.gov.au, not against the version we worked on two years ago or the one the source material was originally written under.
Anthropic Claude
Default large context reasoning model. Used for content rewrite, structural alignment, cross validation passes, and as the reasoning layer inside the custom AI learning app. Reasoning quality matters when the output is going in front of regulated learners.
OpenAI and Google Gemini
Second pass cross validation. Different model, different blind spots. The catch rate on subtle compliance gaps and structural drift goes up when two different large context models review the work independently.
Controlled AI workflow pipelines
Not prompt and hope. Every rewrite runs through a structured workflow with source material checkpoints, output validation, cross model review, and human sign off. The opposite of one shot ChatGPT content that loses Training Package alignment in the first sentence.
Adobe InDesign
Premium V2 visual production layer. Where the workbook deserves to look like a published piece of training material, not a Word document with a logo dropped on the cover. Reserved for V2 builds and multi unit visual systems.
Figma
Design system layer for RTOs running visual identity across multiple units, qualifications, or full Training Packages. The visual layer scales as the unit library grows without each unit getting rebuilt from scratch.
Notion
Content management, version control, draft review, and the layer the team uses to review work in progress without emailing PDFs back and forth for two weeks. Source material, working drafts, and feedback all live in one place.
Custom AI learning app stack
React or React Native front end, Anthropic Claude API as the reasoning layer, PostgreSQL for learner state, hosted on Vercel or AWS depending on the build. The same engineering discipline we bring to custom app development, applied specifically to learner facing tools.
LMS integration patterns
SCORM, xAPI / Tin Can, Moodle, Canvas, and custom internal LMSes. Where the brief calls for direct integration with your existing learning system, we wire it. Where the brief calls for the workbook to sit alongside the LMS without integration, we ship that instead.
Insider RTO review process
Every build gets reviewed by an operator with current RTO delivery and course writing experience. Three and a half years inside a hospitality RTO. Two nationally recognised courses written from scratch. Trade background across mechanical, transmission, and hospitality work. Course material reviewed by someone who has been in the room with real learners.
Sovereign Australian AI compute is on our roadmap. For builds where learner data sensitivity matters more than convenience, we design data handling around that constraint from day one and flag any inference that currently crosses international borders.