Veronica.AI - agentic course advisor

Helps with course discovery according to interest, FCE requirements
Veronica.AI helps you with course advising, no request is too obscure!

Inspiration

We are tired of disparate information and help guides that students use when planning their semester schedule. Our in-person course advisor may not always be available to help us. Hence, we built an LLM-powered agentic course advisor, with access to the latest info sources.

VeronicaAI is aimed at solving the pain points of CMUcourses, SIO, Stellic, and ChatGPT. Inspired by CMU SCS First Year Advisor - Ms Veronica Peet.

Slide deck: https://www.canva.com/design/DAGy3-qcyYc/KKjFrcGUetUEeg2edNK9jA/edit

What it does

Chat with LLM, vibe-plan your semester schedule! Our app understands your taken courses, your major's graduation requirements, and the available courses for the upcoming semester.

How we built it

n8n to orchestrate the agents. We solve qualitative queries with Pinecone RAG, quantitative/structured data queries with an agentic Airtable MCP helper. We also built web scraping tools to fetch graduation requirements and the latest FCE data (inspired by CMU courses' internal data pipelines). Front-end is built with React.

Challenges we ran into

Token rate limits on OpenAI models! We overcome this by using GPT-4.1-mini, instead 200k tokens per minute.

Accomplishments that we're proud of

It can be used easily (talk to it in natural language), and it can pull up the latest information. Besides course advising (recommending courses to take), it can help retrieve the latest information (FCE data, Course Description) about any course and offer comments and evaluations on top.

What we learned

How to do agent orchestration and how to RAG

What's next for Veronica.AI - agentic course advisor

Work on getting FCE data for other schools, e.g., MIT, Stanford, Harvard. And to do QA to make sure it performs well for all majors at CMU / different query types.