Have you ever thought about running AI models locally? It's surprisingly economical, and can be an excellent learning tool.

But is it actually useful? Can you write code with it? Can it handle agentic tasks at least as well as the frontier models?

In this session, Spencer will discuss the current state of running local AI models. We'll discuss model selection, quantization, hardware requirements, performance, and the tools that make local inference possible. There's something for everyone, whether you have a basic laptop or a high powered AI-capable workstation.

We'll close by showing off a real AI cluster - in the very room that the talk is in - and invite the audience to try it themselves!

This will be delivered with 2-4 DGX Sparks as real computers that we'll use to setup and run local models RIGHT IN THE ROOM. At some point the audience will be invited to use a website hosted by and powered by these models! :D