Abstract
Abstract—Access to specialized medical care is uneven across the world. Rural populations, non-English speakers, and low-income patients routinely face delays that worsen outcomes. This paper describes MediVA AI, a web-based healthcare platform that pairs a conversational AI assistant with a secure doctor scheduling backend. A patient first interacts with the AI assistant, which handles preliminary symptom analysis and generates actionable guidance. When the assistant cannot resolve a case, it schedules a consultation with a verified doctor and forwards a conversation summary so the doctor arrives prepared. Medical images and confidential reports are encrypted using a reversible deep learning model, then stored on a blockchain via IPFS hashing—so records are tamper-evident and retrievable without depending on a single server. The system uses Retrieval-Augmented Generation (RAG) with a vector database for context-aware medical responses, a SQL database for structured patient and appointment data, and a NoSQL store for unstructured media. Multilingual support, voice input, and report-upload features make the platform usable across demographics. We describe the architecture, the encryption pipeline, the scheduling logic, and the doctor-facing dashboard. Early testing shows the assistant handles common queries accurately and escalates appropriately when confidence is low. Keywords—AI healthcare; RAG; blockchain; medical image encryption; doctor network; voice assistant; deep learning; IPFS; telemedicine