SparkSprout

AI-powered learning platform that personalizes education for kids through parent-uploaded materials, generating engaging stories and activities with their favorite characters.

Comment

SparkSprout 🌱

Inspiration

When my 7-year-old son—who has high-functioning autism—fell in love with lions, every worksheet, bedtime story, or game that featured lions suddenly held his full attention. I kept wondering:

What if every child's learning materials could pivot around their own fascinations—instantly?

But parents juggle syllabi, sight-word lists, and homework packets; teachers can't customize for dinosaurs, robots, unicorns, and lions all at once. SparkSprout began as the idea of a personalized AI tutor that can ingest any school material and spin it into a learning experience themed for one specific child.

What it does ✨

  1. Upload – Parent drops in a PDF/DOC/image of weekly work.
  2. Extract – AI pulls out spelling words, vocabulary, learning objectives.
  3. Personalize – Child selects an interest and gets assigned a mascot avatar.
  4. Generate – SparkSprout creates personalized stories, worksheets, songs, games, and flashcards that embed those exact words.

The child reads, listens, and enjoys interactive content—without realizing it's homework—while school objectives are seamlessly reinforced.

How we built it

Layer Stack Highlights

Layer Technologies Features
Frontend React 18 + TypeScript (Vite 5.4), Tailwind CSS 3.4, Framer Motion 11 Utility-first design system, smooth page transitions, accessible components
Backend/Supabase PostgreSQL + RLS, Auth, Storage buckets Row-Level Security + JWT, Users only touch their own data
AI OpenAI GPT-4o (content + images), Anthropic Claude (document analysis) Document analysis → content generation pipeline
Document Pipeline PDF.js 3.11, Mammoth.js 1.9 Handles PDF, Word, text, and image formats

Architecture Features

  • Dynamic theme system - Child profiles get personalized mascots with custom color schemes
  • Multi-format support - PDF, DOC, DOCX, TXT, and image file processing
  • Content library - Browse, filter, download, and manage generated materials
  • Child-friendly interface - Separate dashboards for parents and children
  • Security-first - Row Level Security ensures data isolation between families

Challenges we ran into 🔧

Hurdle Fix
Multi-format document processing Integrated PDF.js for PDFs and Mammoth.js for Word docs, with fallback text extraction
Secure file storage Implemented Supabase Storage with organized folder structure (user_id/child_id/filename)
AI service orchestration Built retry logic and rate limiting for reliable Claude → GPT-4o content generation chain
Dynamic theming Created CSS variable system that updates child interface colors based on mascot selection

Accomplishments we're proud of 🏆

  • Complete learning pipeline – Upload → AI Analysis → Content Generation → Interactive Reading
  • Comprehensive content types – Stories, worksheets, songs, games, and flashcards
  • Scalable architecture – React + TypeScript + Supabase foundation ready for growth
  • Child-centered design – Personalized mascots, celebrations, and age-appropriate interfaces

What I learned 🔍

  • Structured AI prompts work better - Using specific content templates and JSON schemas improved generation quality and consistency
  • Security by design - Implementing Row Level Security from the start prevents data access issues as the platform scales
  • Component architecture matters - Well-organized React components (ui/, children/, content/, uploads/) made feature development faster
  • Document analysis complexity - Different file formats require different extraction strategies and validation approaches

What's next for SparkSprout 🚀

Phase 1 (based on current codebase)

  • Progress tracking system - Reading analytics and comprehension metrics
  • Enhanced content library - Better filtering, search, and organization
  • Mobile optimization - Responsive improvements for tablet and phone usage

Phase 2 (planned features)

  • Interactive quizzes - Comprehension testing with immediate feedback
  • Teacher collaboration - Classroom integration and sharing features
  • Advanced accessibility - Screen reader optimization and dyslexia-friendly options

Phase 3 (future vision)

  • Native mobile apps - iOS and Android with offline content access
  • Community features - Parent sharing and collaborative learning
  • Video lessons - Multimedia content generation

Every child will soon be the hero of their own learning adventure—with personalized content that transforms homework into engaging stories & activities tailored to their unique interests.

Built With

Share this project:

Updates