EngStudy
An AI study platform for engineering students — a full RAG pipeline from course PDF to grounded tutor.
HNSW
vector index for retrieval
SM-2
spaced-repetition scheduling
4
study modes from one upload
Upload course PDFs; EngStudy extracts, chunks, and embeds the content, then generates quizzes, SM-2 spaced-repetition flashcards, and study guides — and answers questions through a streaming AI tutor grounded strictly in the uploaded material.
Grounded, not generative-at-random
The design constraint that matters: the tutor answers strictly from the uploaded course material. Documents are chunked and embedded with OpenAI's text-embedding-3-small, stored in PostgreSQL with pgvector, and retrieved through an HNSW index by cosine similarity. Retrieved chunks are the only context the model sees, which keeps answers anchored to the actual course.
Quiz generation, flashcards, and study guides run off the same pipeline — one upload produces four study modes, with quiz attempts and study sessions logged for analytics.
Highlights
- Full RAG pipeline: PDF upload → text extraction → chunking → OpenAI text-embedding-3-small → pgvector with HNSW index → semantic retrieval at query time
- Streaming AI tutor grounded in embedded course material — answers come from the documents, not the model's imagination
- AI-generated quizzes (multiple choice, true/false, short answer), timed and server-graded
- Flashcard decks scheduled with the SM-2 spaced-repetition algorithm
- Study guide synthesis: summaries, formula sheets, exam reviews, and key takeaways in Markdown
- Study sessions and quiz attempts logged for progress analytics