J.I.T. (Just-In-Time) is an AR-powered educational assistant built at LA Hacks 2026 that turns any physical textbook into an interactive 3D learning experience. Built by Anton Ko, Darren Zhao, Timmy Phan and Kenneth Aiden Yandell, J.I.T. won the MLH Best Use of MongoDB Atlas award by bridging the gap between physical textbooks and on-device AI.
What it solves
Textbooks are the foundation of academic knowledge but one of the most static and intimidating ways to learn. Students struggle to stay engaged with dense blocks of text and have no way to ask questions about the material without switching to a search engine that has no context about what they are reading. J.I.T. makes the book talk back.
What it does
Users scan the ISBN barcode of any supported textbook and instantly a 3D avatar named J.I.T. spawns directly on top of the physical book through the phone screen using AR. Students can then have a real-time voice conversation with J.I.T. about the book's content. Because J.I.T. is grounded in the specific knowledge of that textbook using RAG, the answers are factually accurate, course-aligned and far more engaging than a standard search engine.
How it works
J.I.T. is built in Swift with ARKit for the iOS frontend and FastAPI on the backend. A custom RAG pipeline uses MongoDB Atlas as the vector database, converting textbook data into high-dimensional vectors with an embedding model. Cosine similarity retrieves the most relevant context from the textbook to ground the LLM's responses. ElevenLabs handles the real-time voice interaction, making conversations with J.I.T. feel natural and immediate.