LLM Course Designer

I wanted hands-on experience with large language models while also dealing with MSOE's transition from quarters to semesters. That shift forced a lot of instructors to reorganize familiar material into new course structures, which made it a perfect real-world problem.

I built a Gradio-based interface with a Python backend and database support so multiple users could generate and manage content for different courses. The system uses prompt templates built from information entered by the instructor, and users can accept, edit, or replace LLM-generated output at each step.

The tool can generate course modules, expand them into lectures, populate Google Slides, insert relevant images, and export assessments to Canvas. It can also generate lecture videos using text-to-speech from slide transcripts.

Some parts were immediately useful during the semester transition, especially module and quiz generation. Long-form lecture generation was a different story and needed a lot of editing, which pushed my more recent work toward RAG, knowledge graphs, and reflection techniques.