J.A.R.V.I.S: A Multi-Tenant AI Assistant Platform with Retrieval-Augmented Generation, Face Recognition, and Autonomous Browser Automation

This work presents the design and development of J.A.R.V.I.S (Just A Rather Very Intelligent System), a unified multi-tenant AI assistant platform that brings together three independent intelligent technologies within a single web ecosystem. The system integrates a Retrieval-Augmented Generation (RAG) knowledge engine, a real-time facial recognition module for student identification, and an autonomous browser agent capable of executing web- based tasks through a Planner–Executor framework. Unlike conventional AI assistants that operate purely as conversational systems, J.A.R.V.I.S is designed to function as an organizational intelligence layer. It can analyze private institutional documents, identify individuals using biometric encoding, and automate repetitive online workflows. The platform is implemented using a modern full-stack architecture comprising Flask for backend services, React for frontend interaction, FAISS for vector storage, Google Gemini for language reasoning and embeddings, and Playwright for browser automation. To support multi-organization deployment, the system incorporates strict tenant isolation through organization-specific vector collections, role-based access control, JWT-secured authentication, and API- key-based external integrations. Experimental deployment demonstrates that the platform can successfully process heterogeneous documents, generate context-grounded answers with verifiable sources, identify students through facial biometrics, and autonomously perform browser tasks — all within a secure, scalable environment.