Curious about what it really takes to run modern AI without depending on external providers? This session takes a practical look at self-hosting Large Language Models and AI-powered applications. We'll discuss the current open-source ecosystem, infrastructure choices such as using a Mac, Spark or AMD, inference performance (and the cost-quality tradeoff), model orchestration, security considerations, and common pitfalls. Rather than focusing on theory or vendor offerings, the session is built around real-world examples—including several applications running on the Infinity platform—to illustrate how these technologies can be deployed and used in production.