Omega Training: Learning Design Copilot

Inspiration

Training teams often receive essential policy and practice knowledge as long presentations. Converting that material into clear objectives, active learning, assessment, and an LMS-ready package takes specialist time. Omega Training closes that gap while keeping a human author in control.

What it does

An administrator uploads a PowerPoint deck. GPT-5.6 Terra turns it into a structured SCORM 1.2 course, ready for preview, human revision, asset selection, and download. A transparent readiness report checks identity and structure, learning objectives, active learning, assessment, and recap before publication.

How we built it

Omega Training uses Django, Django REST Framework, the OpenAI SDK, python-pptx, and a custom SCORM 1.2 pipeline. GPT-5.6 Terra is the primary model and GPT-5.6 Luna is the fallback. Model output is constrained and then checked by a deterministic local quality layer so reviewers can see why a package is ready.

The training platform existed before Build Week. During the event, Codex helped create the sanitized submission boundary, migrate the learning-design pipeline to GPT-5.6, add the readiness layer, prepare migrations and tests, and document the implementation and verification.

Challenges we ran into

We needed to introduce a more capable model without losing control of latency, cost, or output shape. We moved from mini/nano models to Terra/Luna with explicit reasoning controls. We also addressed author trust by exposing a readiness report and keeping human review in the workflow.

Accomplishments that we're proud of

• A practical PPT-to-SCORM workflow powered by GPT-5.6
• Reviewable learning-design quality checks instead of a black-box score
• A human-in-the-loop editing and preview flow
• A sanitized private submission repository with tests and security checks

What we learned

Generative course creation works best when the model is paired with deterministic checks and a clear review boundary. Strong output is not enough; authors need evidence that objectives, practice, assessment, and recap are present.

What's next

We plan to add richer media generation, more granular author controls, additional LMS validation, and outcome analytics that connect course quality to learner performance.

Built With

• codex