Modern AI applications combine models, serving frameworks, agents, and accelerator software into a complete stack. This hands-on workshop introduces AMD’s ROCm AI stack and demonstrates how to develop and run multimodal AI applications on AMD GPUs. Participants will gain practical experience using enterprise-class AMD GPU infrastructure through the AMD GPU Developer Cloud. During the workshop, attendees will create an OpenClaw-based AI agent, connect it to a model-serving stack powered by vLLM or SGLang, and explore the key components of modern AI systems. The session will also provide an overview of the ROCm ecosystem and highlight opportunities for developers, students, and technology enthusiasts to get involved through code contributions, testing, documentation, tooling, and community engagement.