LLM Agents at the Edge

The goal of the project will be to create a RAG application that leverages LLMs running at the edge. 

This application will be used by students and faculty to quickly find answers to questions related to OSU specific information.  

Objectives


Objectives:

  • Scrape OSU documentation data
  • Organize and enrich scraped data
  • Create sentence embedding using data
  • Store sentence embedding in a Vector database
  • Set up local LLMs
  • Create notebooks to implement agent frameworks
  • Create a simple RAG workflow

Stretch Goals: 

  • Create a backend application to host the LLMs
  • Create a frontend application with a “chatbot" interface

We will leverage the latest tools and workflows to build this application. 

  • LangGraph
  • LangChain
  • HuggingFace transformers
  • Open source LLMs
  • Vector Databases
  • Web scrapers
  • Jupyter Notebooks
  • AI Agent frameworks

This project will require students to learn about these tools and the workflows that are used. 

For more information on these techniques, please see:

Once we have a working prototype, our stretch goals will be to enhance our simple RAG application with more advanced techniques and then bundle that into a frontend chatbot application.

Motivations


As the LLM space is evolving quickly my motivation is to introduce students to the latest technologies and tools to build RAG applications. 

With this application, we could provide a tool for students and faculty to leverage that could provide them ways of quickly finding answers to questions related to OSU through a chatbot like interface. 

Qualifications


Minimum Qualifications:
None Listed

Preferred Qualifications:
None Listed


Details


Project Partner:

Kyle Prouty

NDA/IPA:

No Agreement Required

Number Groups:

1

Project Status:

Accepting Applicants

Keywords:
WebData ScienceMachine Learning (ML)Data EngineeringNew Product or Game
Card Image Capstone