Vector search is often introduced through chatbots and Q&A demos, but that’s only a small part of what it enables.

In this talk, we’ll take a familiar CRUD-based .NET application and incrementally evolve it into something far more interesting: a **multimodal search experience** that understands both **text and images**.

Using MongoDB Atlas Vector Search and Voyage AI’s multimodal embedding models, we’ll build a real, working application that allows users to:

* Find animals using natural language descriptions (e.g. “a large grey animal with a trunk”)
* Click on an image and instantly discover visually similar animals
* Combine semantic similarity with traditional metadata filtering — all in a single database

There are no chatbots, no prompt engineering rabbit holes, and no hand-waving. Instead, we’ll focus on the practical engineering decisions that matter: data modelling, embedding generation, indexing, querying, and how vector search fits naturally alongside existing MongoDB workloads.

By the end of the session, you’ll understand how vector search actually works, when it makes sense to use it, and how you can add genuinely useful AI-powered features to your own .NET applications without rewriting everything or introducing unnecessary infrastructure.