CampusCue: An Offline AI That Turns Confusing College Notices Into Clear Next Steps

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

CampusCue — Turn Confusing College Notices Into Clear Next Steps

What I Built

Have you ever received a 5-page college notice and thought:

"Okay... but what exactly am I supposed to do?"

That was the problem I wanted to solve for a friend.

College students regularly receive scholarship notices, exam circulars, registration announcements, internship opportunities, and administrative PDFs written in formal language. The information is there, but figuring out what actually matters can take a surprising amount of time.

So I built CampusCue.

CampusCue takes a college notice and turns it into a simple action plan:

• 📌 What is this notice about?
• ⏰ What is the deadline?
• 👤 Who is eligible?
• 📄 Which documents are required?
• ✅ What do I need to do?
• ⚠️ Are there any important warnings?
• 🔎 Where in the original document did this information come from?

Instead of simply asking AI to summarize a PDF, CampusCue focuses on the question a student actually cares about:

"What do I need to do?"

The goal was to build something useful for one real person first, rather than trying to build a generic AI product for everyone.

Demo

🚀 Live Demo:

https://campus-cue.vercel.app/

The main workflow is:

Upload notice → Analyze locally → Get action plan → Ask questions → Verify the source

The application is designed around a simple principle:

Don't just summarize the notice. Make it actionable.

Code

💻 GitHub Repository:

https://github.com/HITESHVERMA01/CampusCue

The repository contains the frontend, backend, AI integration, document processing, prompts, and setup instructions.

The project is open source so that the workflow can be inspected, modified, extended, and reused.

How I Built It

CampusCue is built around local open-weight AI, rather than depending on a closed AI API.

Core stack

• React + Vite — frontend
• Tailwind CSS — UI
• FastAPI — backend
• Python — document processing and AI pipeline
• PyMuPDF — PDF text extraction
• Ollama — local model runtime
• Gemma — open-weight AI model

The basic architecture is:

College Notice
      │
      ▼
 PDF / Image Upload
      │
      ▼
 Document Extraction
      │
      ▼
 Page-aware Text
      │
      ▼
 Local Gemma via Ollama
      │
      ▼
 Structured Analysis
      │
      ├── Deadline
      ├── Eligibility
      ├── Documents
      ├── Action Items
      └── Warnings
      │
      ▼
 CampusCue UI
      │
      ├── Action Checklist
      ├── Questions
      └── Source Verification

One of the most important design decisions was keeping page references with extracted information.

If CampusCue says:

"The deadline is October 12."

the user can see where that information came from in the original notice.

This makes the system more useful than a black-box summary and helps reduce the risk of blindly trusting an AI-generated answer.

Why Does Open Innovation Matter?

This is the part of the project that matters most to me.

College notices can contain information that students may not want to send to an external AI service.

A traditional cloud-based approach could look like:

Private Document
      ↓
Internet
      ↓
Third-party AI API
      ↓
AI Response

CampusCue can instead run the core AI workflow locally:

Private Document
      ↓
Student's Computer
      ↓
Ollama
      ↓
Gemma
      ↓
Answer

That changes what is possible.

🔒 Privacy

When using the local model, the document can be processed on the user's own computer instead of requiring it to be sent to a third-party AI API.

💰 No per-request AI API cost

Once the model is available locally, the core inference doesn't require paying for every question or document analyzed through a cloud API.

🔧 Model freedom

The application isn't permanently tied to one proprietary model or provider.

The local model can be replaced and the AI pipeline can be experimented with or extended.

🌐 Offline potential

Because the model runs locally, the core AI functionality can work without an internet connection once the required software and model are installed.

For me, this is the real value of open innovation:

The AI isn't just something I'm calling. It's something I can actually run, inspect, change, and build around.

My Agent Session

Coming soon / [add DevRelay session here]

I also used the build process to explore how an AI-assisted development workflow could help move from an idea to a working product quickly.

Prize Categories

🏆 Best Use of Gemma

CampusCue uses an open-weight Gemma model through Ollama as the core reasoning engine for understanding college notices and turning them into actionable information.

The model isn't simply generating a chatbot response — it is part of the core document-analysis workflow.

---

What's Next?

CampusCue is intentionally small right now.

The next things I'd like to explore are:

• Better OCR for scanned notices
• More robust document retrieval for long PDFs
• Calendar integration for extracted deadlines
• Personalized explanations based on a student's situation
• Support for Hindi/Hinglish explanations
• More reliable citation and evidence extraction
• Fully offline packaging for students

But the most important next step isn't another feature.

It's giving CampusCue to the friend I built it for and seeing whether it actually makes their life easier.

Because that was the point of the challenge in the first place.

Build for one person. Solve one real problem. Then see where it goes.