Studybook LM

I built Studybook LM to rethink how students and researchers interact with fragmented study materials. Inspired by NotebookLM, I set out to build an intelligent workspace where anyone can synthesize diverse sources into a deeply grounded, hallucination-free learning environment.

Under the hood, I developed an end-to-end RAG pipeline supporting PDFs, YouTube transcripts, web articles, and raw markdown notes. Ingested content is cleaned, chunked, and embedded into isolated Pinecone namespaces, enabling real-time streaming chat with strict source citations, rolling conversation summaries, and personalized user memory.

Beyond conversational Q&A, I designed an interactive Studio suite to convert static documents into active recall tools. With a single click, users can generate dynamic flashcard carousels, practice quizzes with detailed explanations, visual concept mind maps, and structured briefing reports tailored directly from their selected materials.

Architecturally, I engineered the application as a scalable TypeScript monorepo leveraging Next.js 16, Express 5, tRPC, and Prisma. To ensure UI responsiveness and prevent serverless timeouts, I offloaded heavy document parsing, vectorization, and artifact creation to Inngest background workers for resilient, distributed task execution.