Inspiration
As B.Tech Data Science students, we constantly struggled with finding concise, personalized answers from huge PDFs and class notes. The idea sparked from our real need — what if we could upload our study material and talk to an AI tutor like it’s a friend?
What it does
AI Study Buddy lets students:
- 📄 Upload study material (like PDFs, notes)
- ❓ Ask questions directly
- 💬 Get smart, context-aware answers
- 🔊 Listen to voice responses via ElevenLabs
It's like having your own AI-powered tutor — available 24/7, in your voice and pace.
How we built it
- 🔗 Used Together AI's Meta-LLaMA 3.1 model for high-quality, free and open-source chat completion
- 🔉 Integrated ElevenLabs Text-to-Speech API to turn answers into clear, natural voice
- ⚙️ Developed with React + TypeScript for the frontend
- 🧠 Backend in Node.js with Vite + Express
- ☁️ Hosted using Netlify, voice streamed through ElevenLabs
- 🧪 Used Bolt’s AI tools to rapidly prototype the app UI/UX
Challenges we ran into
- Managing limited API quotas (OpenAI/Gemini)
- Getting file upload → extraction → chunking → prompt routing to work smoothly
- Adapting large language models to focus only on the uploaded context
- Making the UI simple for users, but powerful behind the scenes
What we learned
- How to work with multiple AI providers efficiently (Together AI, ElevenLabs)
- Handling prompt engineering to get precise educational answers
- Deploying fast MVPs using Bolt’s visual interface
- Turning user feedback into real-time UI improvements
What's next
- Add support for more file formats (PPT, DOCX)
- Bookmark answers / Add to Flashcards
- Voice input (Speech-to-Text)
- Multilingual voice output for regional accessibility
Built With
- bolt
- echowave.io
- elevenlabs.ai
- netlify
- node.js
- react
- taliwindcss
- togetherai
- typescript
- vite
