JARVIS is a hackathon project that builds a real-time people-research system. The idea is that you point a camera at someone and the system automatically identifies who they are and assembles a profile of publicly available information about them within seconds. The original demo used Meta Ray-Ban smart glasses as the camera input, though you can also upload a photo manually. The pipeline works in stages. First, face detection software spots a face in the frame and generates a mathematical fingerprint of it. That fingerprint is sent to a reverse-image search service called PimEyes to identify who the person is. Once a name is found, a group of automated browser agents fans out across LinkedIn, Twitter, Instagram, and Google simultaneously to collect publicly available information. A separate search API called Exa provides a faster first pass. All that raw data is then sent to Claude or Gemini, which synthesizes it into a structured summary called a dossier. The results stream live to a frontend that displays them on a military-style corkboard interface with pushpins and connecting strings. The backend is written in Python using FastAPI. The frontend is built with Next.js. Real-time data syncing across browser tabs uses a service called Convex, though the app also works without it using temporary in-memory storage. All of the external services are optional, the system degrades gracefully when API keys are missing rather than crashing. The project was built for a web agents hackathon organized by Browser Use and Y Combinator. It requires API keys for several paid services including PimEyes, Browser Use cloud sessions, Anthropic, and Exa. The README does not specify a license. Given the use of facial recognition and automated social-media scraping, this is an experimental tool with significant privacy implications.
← affaan-m on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.