StudyAI

• 

  The voice AI Study partner

Inspiration

StudyAI was born from the struggle students face when trying to synthesize complex course materials into actionable study aids. We noticed learners spending more time organizing content than actually studying it. Our inspiration came from combining AI-powered content analysis with interactive learning tools to create a seamless study experience that adapts to individual needs.

What it does

Processes uploaded documents (PDFs, notes) to extract key concepts Generates customized quizzes, flashcards, and study guides by searching relevant content using Orama Provides real-time Q&A through voice and text interaction Offers document-specific tools like content clustering and focus-based question generation

How we built it

Frontend: Next.js with TypeScript, Tailwind CSS, and shadcn/ui AI Services: OpenAI API for embeddings and question generation Database: PostgreSQL with Prisma ORM Realtime Communication: WebRTC for voice interaction Document Processing: Custom PDF parser and Orama for vector search Deployment: Vercel with serverless functions

Challenges we ran into

Algorithm for Creating Questions and Answers that are diverse and also correct and relevant from document Document Processing: Handling large PDFs while maintaining performance.also splitting the document in the right way Real-time Voice: Syncing audio streams with AI responses Question Quality: Ensuring generated questions match learning objectives

Accomplishments that we're proud of

Created a fully functional voice-controlled study assistant Developed a document analysis pipeline that maintains context Implemented a responsive UI that adapts to different study modes Achieved sub-second response times for most queries

What we learned

The importance of chunking strategies for LLM context windows How to balance real-time updates with computational efficiency Techniques for maintaining conversational context across tools The value of progressive UI rendering during long operations Best practices for multimodal (voice+text) AI interactions

What's next for StudyAI

FineTunning the App to Make it suitable for Production and Able to Handle Multiple Content Files. Learning From User Behaviour on how well to instruct the AI. Optimizing to Reduce Cost.

Built With

• neondb
• nextjs
• openai
• orama
• prisma
• trpc