LearnSphere Project Details

📋 Project Overview

LearnSphere is an innovative AI-powered educational platform that transforms how students learn and practice. The platform leverages advanced AI models, particularly GPT-OSS-120B via Groq Cloud API, to provide personalized learning experiences through intelligent problem solving, adaptive learning paths, and progressive exam simulation.

🎯 Inspiration

The inspiration for LearnSphere came from observing the challenges students face in traditional learning environments:

• One-size-fits-all approach: Traditional education doesn't adapt to individual learning paces and styles
• Limited feedback: Students often struggle with problems without getting detailed, step-by-step explanations
• Exam anxiety: Lack of progressive practice leads to poor exam performance
• Language barriers: Educational resources are often limited to specific languages
• Accessibility: Quality tutoring and personalized learning paths are expensive and not widely accessible

We envisioned a platform that could democratize high-quality, personalized education using the power of open-source AI models, making advanced learning assistance available to everyone regardless of their location, economic status, or native language.

🚀 What it does

LearnSphere provides three core AI-powered educational features:

1. 🧠 AI Solver

• Multi-format Problem Analysis: Students can upload images, PDFs, or type problems directly
• Step-by-step Solutions: GPT-OSS-120B generates detailed explanations with mathematical equations
• Multi-subject Support: Covers Math, Physics, Chemistry, and Biology
• Multi-language Support: Provides solutions in 5 languages (English, Spanish, Chinese, French, Vietnamese)

2. 🗺️ Learning Road

• Personalized Curriculum: GPT-OSS-120B creates custom learning paths based on student goals
• Progressive Structure: 5-8 steps from beginner to advanced levels
• Interactive Content: Rich learning materials with examples, exercises, and quizzes
• Adaptive Difficulty: Content adjusts based on student performance

3. 📝 Exam Simulator

• Progressive Difficulty System: 7 levels from Foundation to Genius
• Adaptive Testing: AI analyzes performance and adjusts subsequent exam difficulty
• Performance Analytics: Detailed insights and improvement suggestions
• Real Exam Preparation: Simulates actual exam conditions with time limits and scoring

🏗️ System Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        LEARNSPHERE ARCHITECTURE                 │
└─────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────┐
│                         FRONTEND LAYER                          │
├─────────────────────────────────────────────────────────────────┤
│  React 18 + TypeScript + Tailwind CSS + Vite                    │
│                                                                 │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐              │
│  │ AI Solver   │  │Learning Road│  │Exam Simulator│             │
│  │ Component   │  │ Component   │  │ Component   │              │
│  └─────────────┘  └─────────────┘  └─────────────┘              │
│                                                                 │
│  ┌─────────────────────────────────────────────────────────────┐│
│  │           Language Context (5 Languages)                    ││
│  └─────────────────────────────────────────────────────────────┘│
│                                                                 │
│  ┌─────────────────────────────────────────────────────────────┐│
│  │         Local Storage (Dexie/IndexedDB)                     ││
│  │  • Exam Series  • Learning Paths  • User Progress           ││
│  └─────────────────────────────────────────────────────────────┘│
└─────────────────────────────────────────────────────────────────┘
                                    │
                                    │ Call
                                    ▼
┌─────────────────────────────────────────────────────────────────┐
│                      SERVICE LAYER                              │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐  │
│  │  SolverService  │  │LearningRoadSvc  │  │ExamSimulatorSvc │  │
│  │                 │  │                 │  │                 │  │
│  │ • File Upload   │  │ • Path Gen      │  │ • Progressive   │  │
│  │ • OCR Process   │  │ • Content Gen   │  │   Exams         │  │
│  │ • Step-by-Step  │  │ • Quiz Gen      │  │ • Analytics     │  │
│  └─────────────────┘  └─────────────────┘  └─────────────────┘  │
│                                                                 │
│  ┌─────────────────────────────────────────────────────────────┐│
│  │                   ModelService                              ││
│  │  ┌─────────────┐  ┌─────────────┐                           ││
│  │  │   Gemini    │  │  GPT-OSS    │                           ││
│  │  │   2.0 Flash │  │   120B      │                           ││
│  │  │             │  │ (Groq API)  │                           ││
│  │  │ • OCR       │  │ • Problem   │                           ││
│  │  │             │  │   Solving   │                           ││
│  │  │             │  │ • Learning  │                           ││
│  │  │             │  │   Paths     │                           ││
│  │  │             │  │ • Exam Gen  │                           ││
│  │  └─────────────┘  └─────────────┘                           ││
│  └─────────────────────────────────────────────────────────────┘│
└─────────────────────────────────────────────────────────────────┘
                                    │
                                    │ API
                                    ▼
┌─────────────────────────────────────────────────────────────────┐
│                      EXTERNAL AI SERVICES                       │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  ┌─────────────────┐  ┌─────────────────┐                       │
│  │   Groq Cloud    │  │  Google AI      │                       │
│  │                 │  │                 │                       │
│  │ GPT-OSS-120B    │  │ Gemini 2.0      │                       │
│  │                 │  │ Flash           │                       │
│  │ • High-perf     │  │                 │                       │
│  │   inference     │  │ • Multimodal    │                       │
│  │ • 65K tokens    │  │ • Fast response │                       │
│  │ • Reasoning     │  │ • Vision + Text │                       |
│  └─────────────────┘  └─────────────────┘                       │
└─────────────────────────────────────────────────────────────────┘

🔧 How we built it

Technology Stack

Frontend:

• React 18 + TypeScript: For type-safe, modern UI development
• Tailwind CSS: For responsive, utility-first styling
• Vite: For fast development and optimized builds
• Dexie (IndexedDB): For client-side data persistence
• React Context API: For state management

Backend:

• Python FastAPI: For high-performance API development
• Docling: For advanced document processing and analysis

AI Integration:

We chose GPT-OSS-120B as our primary AI model for several key reasons:

1. Open Source Philosophy: Aligns with our mission to democratize education
2. High Performance: 120B parameters provide excellent reasoning capabilities
3. Cost Effective: Via Groq Cloud API, we get enterprise performance at reasonable costs
4. Fast Inference: Groq's hardware acceleration provides sub-3-second responses
5. Large Context: 65K token limit handles complex educational content

Implementation Details

1. AI Solver Implementation:

// GPT-OSS processes uploaded problems with detailed prompts
const prompt = `You are an expert tutor. Analyze this problem and provide step-by-step solutions...`;
const response = await getOssResponse(prompt);

2. Learning Road Generation:

// GPT-OSS creates personalized 5-8 step learning paths
const learningPath = await getOssResponse(createLearningPathPrompt(topic, language));

3. Exam Simulator:

// GPT-OSS generates progressive difficulty exams (7 levels)
const exam = await getOssResponse(createExamSeriesPrompt(description, level));

🚧 Challenges we ran into

1. AI Model Integration Complexity

• Challenge: Integrating multiple AI models with different APIs and response formats

2. Response Consistency and Parsing

• Challenge: AI models sometimes returned inconsistent JSON formats or included markdown
• Solution: Implemented robust response cleaning and parsing with fallback mechanisms

3. Multi-language Support

• Challenge: Ensuring AI responses maintain quality across 5 different languages
• Solution: Developed language-aware prompting strategies and validation systems

🏆 Accomplishments that we're proud of

1. Successful GPT-OSS-120B Integration

• Achieved seamless integration with Groq Cloud API
• Maintained sub-3-second response times for complex problems
• Implemented robust error handling and fallback mechanisms

2. Comprehensive Educational Platform

• Built three distinct but integrated learning features
• Created progressive difficulty system (7 levels from Foundation to Genius)
• Implemented real-time performance analytics and improvement suggestions

3. Multi-language Accessibility

• Successfully deployed in 5 languages with consistent quality
• Maintained educational accuracy across different linguistic contexts
• Created language-aware AI prompting strategies

4. User Experience Excellence

• Designed intuitive, responsive interface
• Implemented real-time feedback and progress tracking
• Created seamless workflow across all features

📚 What we learned

1. AI Model Selection is Critical

• Open-source models like GPT-OSS-120B can match proprietary models in educational tasks
• Groq's hardware acceleration significantly improves user experience

2. User Experience in AI Applications

• Response time is crucial - users expect near-instant feedback
• Error handling and fallbacks are essential for production AI applications
• Visual feedback during AI processing improves perceived performance

5. Technical Architecture Lessons

• Modular service architecture enables easy AI model swapping
• Client-side storage is crucial for educational applications

🚀 What's next for LearnSphere

Short-term Goals (3-6 months)

1. Enhanced AI Capabilities

   • Fine-tune GPT-OSS-120B on educational datasets
   • Implement adaptive learning algorithms
   • Add voice interaction capabilities

2. Expanded Subject Coverage

   • Add Computer Science and Programming
   • Include History and Literature
   • Expand to K-12 curriculum alignment

3. Advanced Analytics

   • Implement learning analytics dashboard
   • Add predictive performance modeling
   • Create personalized study recommendations

Novelty of the Idea ⭐⭐⭐⭐⭐

Creative Innovation:

• Progressive AI Difficulty: First platform to implement 7-level AI-generated progressive difficulty
• Multi-modal AI Education: Unique combination of problem solving, learning paths, and exam simulation
• Open-Source AI Education: Pioneer in using open-source LLMs for comprehensive education

Existing Solutions Comparison:

• Khan Academy: Limited to pre-recorded content, no personalized AI tutoring
• Chegg/Course Hero: Provides answers but not step-by-step learning paths
• Duolingo: Language-specific, doesn't cover STEM subjects
• Traditional Tutoring: Expensive, limited availability, not scalable

Improvements Over Existing Solutions:

• Real-time AI Tutoring: Instant, personalized explanations
• Comprehensive Learning: Combines solving, learning, and testing
• Global Accessibility: Multi-language, cost-effective, available anywhere
• Open Source: Transparent, community-improvable, aligned with educational values

LearnSphere represents the future of education: AI-powered, globally accessible, and fundamentally designed to benefit all of humanity through the democratization of high-quality, personalized learning experiences.

Built With

• gpt-oss
• groq
• react