TBR AI EXCHANGE

AI Learning Collaborative

Agentic Content Creation from Notes

Submission Date

Submitter’s Name/Email

  • Mike
  • Knap
  • michael.knap@nscc.edu

Institution/School

  • Nashville State Community College (NSCC)

Department/Discipline

  • Discipline-Specific Faculty – STEM & Technical Fields

Activity Purpose (assessment, data collection, classroom management, etc.)

  • Accessibility/UDL, Brainstorming/idea generation, Collaboration, Course design/delivery, Content creation/updates, Meeting notes / summarization, Productivity, Personalized learning / teaching

AI Tool(s)

  • Open source, python, open-weight models, custom code, multimodal AI input and output

Activity Details

AI-Supported Math Content Creation Workflows

I have created a demo website of several workflows and generative capabilities of AI to support math content creation. Notably, one workflow can ingest a single handwritten mathematical note and output an accompanying typeset document, with accessibility features that can be added, as well as a narrated video working through the notes. The narration can be personalized in other languages and voices.

The flow provides faculty with origin and ownership of content. The supporting documents only complement the original work. The generative work must be reviewed by the author. The work is, however, supplied as easily editable LaTeX and Python files for improvement and refinement.

Moreover, this flow is AI-agnostic. Frontier models from any provider may be used. I, however, experiment with local models. The attached files demonstrate the workflow with a local AI model, Gemma 4-12B. I have attached the input document and the unrefined, raw output from the model. This content was produced in about one minute with a local model. I have produced a narrated video and slide to accompany this; however, I am not allowed to upload them here.

It is evident that there is room for improvement; this is the work I would pursue in the Fellowship.

Comments

There is so much work being done in the open source community where massive improvements of this kind of workflow are being developed, like using multiple model orchestration, tool calling and development and quality assurance loops. This is only the beginning of what can be possible with AI tooling.

I would be happy to provide more information on my website and examples
Back to AI Use Cases