The system
I developed a workflow where an LLM model generates academic content, which is then processed with Python to ensure scientific consistency and automatically typeset in LaTeX.
This approach makes it possible to:
- Generate unique variants of exams and exercises.
- Maintain professional scientific typographic quality.
- Automate grading through metadata embedded directly in the LaTeX source code.
The orchestration engine for this project is n8n, which connects language model requests with the local rendering server.