AI-powered standardized test prep platform (GRE & GMAT) — question generation engine, web app, and mobile app (coming soon).
CompEx-Platform/
├── web/ — Next.js + Prisma web application (hosted on Vercel, Neon DB)
├── qgen/ — Python AI question generation engine (Deepseek LLM, Docker)
├── mobile/ — React Native / Expo mobile app [coming soon, same Neon DB]
├── planning/ — Architecture docs, import specs, product roadmap
└── README.md
The student-facing platform built with Next.js, TypeScript, and Prisma ORM.
- Live: Hosted on Vercel
- Database: Neon DB (PostgreSQL — shared with mobile)
- Serves GRE & GMAT practice questions, tracks student performance, and powers the AI tutoring experience
An AI-powered Python pipeline that auto-generates complete standardized test papers.
- LLM: Deepseek (primary), Gemini (fallback)
- Parallelism: Up to 105 concurrent API threads (21 sections × 5 threads)
- Output: ~1,849 questions across GRE Quants, GRE Verbal, GMAT Quants, GMAT Verbal, GMAT Data Insights
- Containerized: Full Docker + docker-compose setup
- Questions are imported into the Neon DB via Prisma for the web and mobile apps to consume
A React Native / Expo mobile experience connected to the same Neon DB as the web app.
- Shared database schema, shared question bank, shared student data
- Planned feature parity with the web platform
qgen/ (Deepseek AI generates questions)
↓ JSON output (1,849+ questions)
↓ Prisma import script
Neon DB (PostgreSQL — single source of truth)
↓ ↓
web/ (Vercel) mobile/ (coming soon)
See planning/ for:
QGen-FRESH-Import-Plan.md— Technical spec for migrating generated questions into the DBfuture-concepts/compex-bot-evolution-concept.md— Elo-style AI tutor bot evolution roadmap (v1→v2→v3)
The original standalone repositories are preserved as read-only archives on GitHub:
TanmayKumar-EngStud/CompEx— original web app repoTanmayKumar-EngStud/QGen-py-compex— original question engine repo