This guide explains how to build a local healthcare AI assistant with MedGemma, Google’s open medical model for text and image understanding. The model runs with Ollama for local management and Open WebUI for a ChatGPT-style interface, so you can chat with it and upload medical images such as chest X-rays without sending data to the cloud.

MedGemma is a developer model, not a medical device, so its outputs should not be used directly for clinical diagnosis, patient management, or treatment decisions. It is intended for learning, prototyping, and research, with qualified healthcare professionals needed for real medical questions.

Running the models locally helps protect privacy, avoids per-token costs, supports offline use, and gives you full control over model versions. The 4B model needs at least 8 GB of RAM, while the 27B model requires 32 GB of RAM or a 24 GB+ GPU.