PathFinder

Inspiration

First-generation college students are at a massive disadvantage — not because they're less capable, but because they lack the informal knowledge that comes from having family members who attended college. They miss FAFSA deadlines, don't know free tutoring exists, and never find out about emergency funds when they're struggling. We built PathFinder to close that gap.

What it does

PathFinder is an AI-powered campus navigator that acts as a knowledgeable friend for first-gen students at UMD. Students can:

• Ask any question about campus resources in plain English
• Browse resources by category (Financial Aid, Academic, Career, Wellness, Housing)
• Get personalized guidance based on their year and challenges
• Find specific office locations, phone numbers, and deadlines instantly

How we built it

• Frontend: React 18 + Vite with a custom dark UI
• AI: Llama 3.2 running locally via Ollama
• Backend: Node.js + Express proxy server
• Key feature: A detailed system prompt encoding real UMD-specific knowledge — office names, building locations, free resources, and deadlines ## Challenges we ran into
• Getting the AI to run fully locally without any API costs
• Building a proxy server to connect the React frontend to Ollama
• Encoding enough UMD-specific knowledge into the system prompt to make responses genuinely useful, not generic ## Accomplishments that we're proud of

What we learned

How to run open-source LLMs locally with Ollama, how to design AI systems with real domain knowledge, and how to build a full-stack AI app from scratch in 90 minutes.

What's next for PathFinder

• Live UMD API integrations for real-time deadlines
• Multi-language support (Spanish, Arabic, Chinese)
• FAFSA deadline push notifications
• Saved resource bookmarks and conversation history

Built With

• 3.2
• express.js
• llama
• node.js
• ollama
• react
• vite