EduForge: AI-Powered Educational Resource Generation

An open-source pipeline that generates K-12 educational materials using LLMs and global curriculum standards.

Introduction

I recently released EduForge, an open-source educational resource generation system. The project tackles a problem I've been thinking about for a while: how can we generate high-quality K-12 educational materials at scale, while ensuring they meet global curriculum standards?

What It Does

EduForge ingests high-quality knowledge sources (Wikipedia Vital Articles, OpenStax textbooks, CK-12 FlexBooks) and uses LLMs to generate educational resources: info sheets, worksheets, quizzes, and vocabulary builders. The system targets K-12 education (ages 5-18) and maps to a universal superset curriculum that meets or exceeds the highest global standards.

Architecture

The pipeline consists of several stages:

  1. Curriculum Builder: Parses curriculum standards from Singapore MOE, UK National Curriculum, US Common Core, and others into a unified topic tree
  2. Source Scraper: Fetches and processes Wikipedia articles and open textbooks
  3. RAG Pipeline: Embeds content and retrieves relevant passages for each topic
  4. Resource Generator: Uses LLMs (via NVIDIA NIM) to generate contextually appropriate materials
  5. Quality Judge: Scores generated content on accuracy, age-appropriateness, and curriculum alignment
  6. FastAPI Backend: Serves resources via a REST API with SQLite catalog
  7. React Frontend: Built with Vite, React 19, Radix UI, and Tailwind CSS

Universal Curriculum Framework

Instead of picking one country's curriculum, EduForge builds a superset by taking the highest standard for each subject-grade intersection:

  • Mathematics: Singapore MOE + Shanghai (mastery-based, CPA approach)
  • Science: Finland + UK National Curriculum + NGSS (inquiry-based learning)
  • English Language Arts: UK National Curriculum + Common Core (literature + analytical reading)
  • Social Studies: IB + Cambridge IGCSE (internationally-minded, critical source analysis)

Why It Matters

The system is designed for resource-constrained environments where teachers may not have access to quality materials. All generated resources are designed to be printed and photocopied, making them practical for schools with limited technology infrastructure.

Tech Stack

  • Backend: Python, FastAPI, SQLite, OpenAI SDK (NVIDIA NIM), sentence-transformers
  • Frontend: React 19, Vite, Radix UI, Tailwind CSS, Wouter for routing
  • Infrastructure: Docker, nginx, docker-compose

The system is free, open-source, and designed to work offline after initial content generation.

Status

The project is in active development. Check out the GitHub repository for the full codebase and documentation.