Vidya Setu

🌟 Inspiration

India has 260M+ students across 22+ languages, yet quality education remains inaccessible to many—especially disabled and underserved learners.
Vidya Setu uses AI to break language and accessibility barriers in education.

🚀 What it does

An AI-powered platform that delivers:

• 📊 Adaptive quizzes based on performance
  
• 🌍 Multilingual support (10 Indian languages)
  
• ♿ Accessibility profiles (ADHD, dyslexia, Hearing impairment, Visual impairment)
  
• 📄 PDF learning with a RAG chatbot
  
• 🔊 Text-to-speech for better engagement
  

🏗️ How we built it

• ⚙️ Databricks AI pipeline
  
• 📚 RAG for document understanding
  
• 🤖 Llama 3.3 70B (responses)
  
• 🌐 M2M100 (translation)
  
• 🔊 Sarvam AI (TTS)
  
• ⚡ FAISS + embeddings (retrieval)
  
• 🎨 Streamlit frontend
  

⚠️ Challenges

• ⚡ Accuracy vs latency trade-offs
  
• 🌐 Imperfect multilingual translations
  
• 📄 Complex PDF parsing
  
• 🔄 Streamlit state management
  

🏆 Accomplishments

• ✅ Real-time support for 10 languages
  
• ✅ Adaptive ML-based learning system
  
• ✅ Accessibility-first AI design
  
• ✅ Fully functional RAG pipeline
  

📚 Learnings

• Personalization improves engagement
  
• Latency is critical for UX
  
• Multilingual AI needs optimization
  
• Deployment & scalability matter
  

🔮 What's next

• 🎤 Voice input
  
• 📝 Exam mode + analytics
  
• 🎮 Gamification
  
• 📱 Mobile app + teacher dashboard
  

Built With

• ai
• catalog
• databricks
• deltalake
• embeddings
• faiss
• llama
• model
• model-serving
• pypdf
• python
• rag
• sarvam
• search
• sentence
• serving
• streamlit
• text-to-speech
• transformers
• tts
• unity
• vector
• vector-search