At REA we built our own multimodal AI model to detect nuanced property attributes by combining property descriptions with images. But what does “training our own AI model” really mean? In many ways it is like raising a mischievous dachshund puppy, full of potential but requiring patience, guidance and the right rewards.

In this session you will learn how multimodal models are taught to understand both what they see and what they read. We will break down core concepts such as model selection, embeddings, feedback loops and evaluation metrics in a way that is simple, practical and memorable, told through the story of training my naughty puppy. No deep AI expertise is required. You will leave with a clear mental picture of how machines can learn from words and images, along with ideas you can apply in your own projects.