About the Project — ShikhokAI (OpenShikkhok)

Motivation

Bangladesh's NCTB curriculum reaches millions of students, but there's no real digital tool built specifically around it — most platforms either ignore it or bolt on generic content that doesn't match what students are actually studying for. I wanted to build something that takes the actual curriculum and turns it into active, AI-assisted learning: auto-generated flashcards, a smart digital notebook, and math help that understands handwritten work.

What It Does

• A React/TypeScript SPA where students navigate the NCTB curriculum chapter by chapter
• AI-generated flashcards with a three-tier difficulty system, built from structured chapter summaries (context.json) instead of plain RAG, for much better curriculum alignment
• A Konva.js-based digital notebook storing handwritten strokes as JSONB, paired with AI vision that reads handwritten math and recommends the next concept or step
• A FastAPI backend with async SQLAlchemy, Neon PostgreSQL, and JWT auth

How I Built It

I used AI assistance for most of the actual implementation — architecture decisions, debugging, the flashcard generation rewrite, and the notebook/stroke-storage logic were all built with heavy AI collaboration rather than written from scratch line-by-line. My role was steering the architecture, curating curriculum content, and validating that AI-generated output was actually correct and useful for real students.

Built With

• fastapi
• neon
• openai
• python
• react
• sqlalchemy
• typescript