Case study · 01
PrepXpert — AI-powered test preparation
Overview
End-to-end platform for MDCAT and university entry tests in Pakistan: secure accounts, structured question banks, and AI-assisted practice tuned for high-stakes admissions workflows.
Project preview
Technologies used
- Next.js
- PostgreSQL
- Prisma
- JWT
- OpenAI APIs
Key features
- Adaptive practice flows with streaks, milestones, and lightweight gamification
- JWT-backed authentication with role-aware surfaces for students and admins
- OpenAI-backed explanations and weak-area signals layered into study sessions
- PostgreSQL + Prisma schema oriented toward analytics and long-term growth
- Next.js App Router shell with cohesive UX across marketing and application areas
Challenges solved
- Keeping AI calls responsive and cost-aware under concurrent practice traffic
- Modeling secure, multi-role access without slowing down the student experience
- Unifying product, content, and progress data behind a maintainable service boundary
Architecture & engineering highlights
Edge-aware routing into authenticated handlers, Prisma-backed Postgres for domain state, and a thin AI boundary so prompts and model calls stay isolated from core business logic.
Engineering highlights
Adaptive practice
Structured sessions that emphasize weak topics while preserving predictable latency for users on everyday networks.
Retention design
Progress and streak mechanics aligned with exam timelines so motivation stays tied to measurable outcomes.
Operational clarity
Schema and API choices that make it straightforward to extend question types and reporting as the product grows.
Screenshots
Future improvements
- Deeper analytics on question-level difficulty calibration across cohorts
- Offline-tolerant practice packs with sync when connectivity returns
- Expanded localization for regional syllabi beyond the current entry-test set