AfterBoards + AfterGrad
Two EdTech platforms built solo from zero, covering web, mobile, AI pipelines, admin tooling, and real product operations.
Problem it solved
IPMAT/CUET students had no reliable exam practice platform. Manual exam creation took multiple days per test, which created a hard ceiling on content output and slowed the team every time a new exam had to be prepared.
What I built
I built two full EdTech platforms:
- AfterBoards for IPMAT/CUET students
- AfterGrad for CAT preparation
Each platform included:
- a Next.js web app
- an Expo React Native mobile app shipped to both the App Store and Play Store
- a Server Actions backend
- a PostgreSQL + MongoDB + Redis data layer
- admin workflows for content and operations
The project was not a marketing site with a dashboard attached. It was a full operating product with user accounts, test-taking flows, performance tracking, mobile release management, payments, and internal tooling.
Technical decisions made
Multi-provider AI pipeline for exam generation
The biggest leverage point was exam creation. I built a multi-provider AI system that used OpenAI, Anthropic, and Groq depending on the task and failure mode. The pipeline generated structured drafts, validated them, and handed the result to review in a much more usable format.
Real-time multiplayer word game using WebSockets
To improve retention and make vocabulary practice less static, I built a real-time multiplayer word game. WebSockets handled session updates, turn state, and lightweight presence without dragging the whole platform into unnecessary complexity.
Shared auth across web and mobile with per-device limits
Authentication had to work reliably across surfaces and remain manageable operationally. I implemented an auth system that handled web and mobile access, tracked device usage, and enforced per-device limits without making the sign-in experience feel heavy.
Turborepo monorepo managing 4 apps and shared packages
The system lived in a Turborepo monorepo with four apps and shared packages. That made it much easier to centralise types, validation rules, and reusable logic while still keeping product-specific code from bleeding into everything else.
Outcome and metrics
- 1000+ registered users
- 2000+ exam attempts
- Exam creation reduced from multiple days to under 30 minutes
Those metrics mattered because they reflected actual operational change, not vanity traffic. The platform shipped, people used it repeatedly, and internal content workflows became materially faster.
Tech stack used
- Next.js
- React Native
- TypeScript
- PostgreSQL
- MongoDB
- Redis
- OpenAI
- Anthropic
- Vercel
- Cloudflare R2
- Razorpay