J.A.R.V.I.S. @ Stanford TreeHacks

At TreeHacks 2025 at Stanford, my team and I spent 36 hours hacking together J.A.R.V.I.S.—an intelligent AR assistant designed to enhance human connection and memory using live information retrieval alongside computer vision. Driven by our mission to empower individuals facing memory challenges, such as dementia patients or people with disabilities, we developed J.A.R.V.I.S. to seamlessly integrate real-time facial recognition, voice transcription, and agentic AI reasoning within Apple's Vision Pro headset. Leveraging advanced technologies like Retrieval-Augmented Generation (RAG), LangChain frameworks, Whisper for multilingual transcription, facial recognition models (Dlib), and optimized inference engines like Groq, we built a contextually aware, multimodal assistant capable of recalling past conversations, instantly retrieving personalized information, and intuitively supporting social interactions. Using J.A.R.V.I.S. is transformative — just put on the headset, and the next time you meet someone, our computer vision model will detect their face from the web, obtain the relevant details (LinkedIn, web search) and automatically display that information when you are talking to them. Further, the next time you meet someone, J.A.R.V.I.S. automatically recognizes this and fetches an AI summary of your previous conversation.