LearnFlow - Hackathon Submission

Inspiration

LearnFlow was inspired by the widening educational inequality observed during the pandemic. While some students benefited from personalized tools and private tutoring, millions relied on generic, one-size-fits-all content with no adaptability.

This led us to ask:

What if AI could democratize access to personalized education?

Three key observations strengthened our vision:

Teachers spend 10+ hours per week creating learning materials and assessments
Students forget 70% of newly learned content within 24 hours without reinforcement
Over 1.2 billion students worldwide lack access to personalized learning support

LearnFlow exists to ensure that every learner has access to a personalized AI tutor, and every educator has tools that amplify — not replace — their impact.

What It Does

LearnFlow is an AI-powered adaptive learning platform that converts any text into personalized learning materials within seconds.

For Students

Instant Study Material Generation: Upload or paste content to generate quizzes, summaries, and flashcards
Smart Spaced Repetition: Implements the SM-2 algorithm to optimize review timing
Adaptive Feedback: Prioritizes weak topics and adjusts difficulty based on performance
Progress Analytics: Tracks mastery, recall patterns, and study behavior
Gamification: Includes streaks, badges, and learning milestones
Multi-language Output: Supports English, Spanish, and French

For Teachers

Class and Student Management
AI-Generated Assessments with difficulty tuning
Student Progress Dashboard
Performance Insights and Alerts
Reduced Content Creation Workload

The Technology

AI content generation using Mistral AI
Custom implementation of the SM-2 spaced repetition algorithm
Real-time analytics with adaptive difficulty
Accessible and intuitive UI inspired by modern learning platforms

How We Built It

Frontend

React 18 + Vite
Axios
Recharts for visual analytics
Tailwind CSS for responsive UI
React Router and Lucide Icons

Backend

FastAPI (Python)
SQLite + SQLAlchemy ORM
JWT authentication and Bcrypt for hashing
Pydantic for validation
Mistral AI integration

Algorithms and Architecture

Full SM-2 spaced repetition logic
Adaptive learning rules based on performance trends
Analytics engine using weighted scoring and indexed database queries

Development Approach

Research cognitive science methods and existing learning systems
Design modular architecture with async API operations
Build core learning and assessment features
Layer advanced features: spaced repetition, analytics, and dashboards
UI polish, user flows, and error handling
Testing and deployment optimization

Technical Highlights

Fully working spaced repetition learning flow
Real-time progress tracking with optimized database queries
Clean, maintainable, and scalable API design
Cross-platform friendly deployment adjustments

Challenges We Faced

1. AI Response Consistency
Mistral’s responses varied in structure. We solved this with stricter prompts, schema validation, and fallback parsing.

2. Spaced Repetition Edge Cases
Handling first-time cards, failed reviews, and ease factor adjustments required extensive testing and iteration.

3. Efficient Progress Analytics
Tracking trends for potentially large datasets required query optimization and caching.

4. Deployment Compatibility
Platform differences caused build permission issues, resolved by updating run commands for portability.

5. Balancing Simplicity vs. Power
We refined the UI with role-based views and progressive feature reveal.

Accomplishments We’re Proud Of

A complete, functional platform — not just a prototype
Implemented a scientifically proven memory optimization technique (SM-2)
Built a design experience that is both modern and accessible
Delivered meaningful features for both educators and students
Strong technical execution with secure authentication, analytics, spaced repetition, and AI generation