CollegeCompass-AI

Inspiration

The college application process is fragmented and stressful. Students need to track university requirements, deadlines, costs, financial-aid policies, and personal fit across many websites—often without knowing which information is current or relevant to them. We wanted to build a tool that turns this research into a clear, personalized plan rather than another generic AI chat.

What it does

CollegeCompass AI is a source-aware college admissions workspace. Students can create College Collections, research universities through official pages, review sourced requirements and deadlines, compare fit, ask AI questions using selected parts of their profile, and turn recommendations into actionable tasks and milestones.

The app keeps sourced facts separate from AI interpretations and includes citations, helping students verify important information before acting on it.

How we built it

We built the project with Next.js 16, React 19, TypeScript, Supabase, Featherless AI, and optional Tavily search.

Supabase manages authentication and persistent data including student profiles, college collections, research findings, requirements, AI results, citations, and tasks. We use row-level security to protect student-owned data.

For AI, we use different model roles: smaller models for routing and structured extraction, a reasoning model for fit analysis and comparison, and deeper reasoning only for complex or conflicting evidence. Outputs are validated before being stored.

Challenges we ran into

One challenge was making AI useful while keeping it trustworthy. Admissions details can change, and incorrect information can have real consequences. We addressed this by restricting research to verified university domains, attaching sources to findings, storing facts separately from AI analysis, and clearly surfacing uncertainty.

We also had to balance personalization with privacy. Students may want academic-fit advice without sharing essays or financial details. Our Student Data Scopes let users choose exactly what profile information an AI action can access.

Accomplishments that we're proud of

We are proud that CollegeCompass AI is more than a chatbot. It is a structured workflow that helps students move from research to action:

Choose a goal → Select a College Collection → Select a data scope → Run an AI action → Review and save or create tasks

We are also proud of the source-aware design, privacy controls, persistent research workspace, and ability to convert recommendations into practical next steps.

What we learned

We learned that trustworthy AI experiences require more than a capable model. Good AI products need reliable data boundaries, validation, citations, and interfaces that help users make decisions.

We also learned that using the smallest reliable model for each task improves efficiency and maintainability. Extraction, routing, and complex reasoning have different needs, so assigning them to the right AI role makes the system more practical.

What's next for CollegeCompass-AI

Next, we want to expand the university database, add more automated requirement and deadline refreshes, improve personalized fit analysis, and introduce richer application timelines and notifications.

We also plan to add stronger college comparison tools, more detailed financial-aid support, and collaborative features so students can share progress with counselors, parents, or mentors while maintaining control over their data.

Built With

• and
• featherless-ai
• optional
• react-19
• supabase
• tavily
• typescript