Collexa — College Decision Intelligence Platform

Problem

With thousands of colleges, shifting fees, and confusing rank cutoffs, students end up piecing together decisions from scattered PDFs, forums, and outdated rank-predictor sites.

Solution

A Next.js App Router monolith (TypeScript, Prisma, PostgreSQL) that lets students search & filter colleges, run a weighted rank-based admission predictor, compare shortlists side by side, and get AI-assisted guidance — all behind Zod-validated APIs and HTTP-only JWT cookie auth.

Features

• Search & filter across colleges by state, type & fees
• Weighted rank-prediction engine (rank, budget & placement scoring)
• Side-by-side college comparison tool
• AI guidance chat via a server-side Groq API proxy
• JWT auth in HTTP-only cookies + middleware-enforced access control
• Zod-validated API routes with Prisma ORM over PostgreSQL
• Email-based password reset flow (Resend) with expiring tokens

Architecture

• Next.js App Router as the single composition layer — src/app/api holds route handlers colocated with the pages that call them
• middleware.ts centralizes access control, checking the JWT cookie before requests reach any protected route handler
• Relational schema (Prisma): User, College, Course, Review, SavedCollege (unique userId+collegeId), PasswordResetToken
• Recommendation weights (rank/fees/placement) are environment-configurable, not hardcoded

Technology Stack

Challenges

• Indexed lookup fields on college name, location, rating, and user-saved records for fast filtered reads
• Cascade-safe deletes on Course/Review so removing a college doesn't orphan child records
• Global Prisma Client reuse in development prevents connection-pool churn on hot reload
• Password reset tokens are single-use and expiring — no long-lived reset links

Results / Impact

Links