AMD-CineMatch: Mood-Powered AI Agent

AMD-CineMatch addresses the "decision fatigue" of modern streaming by implementing a sophisticated, decoupled AI architecture. Developed during the AMD Developer Hackathon, the project utilizes a high-performance backend powered by the AMD AI Stack (ROCm 6.0) to handle complex metadata analysis and sentiment mapping.The system architecture is split into two primary layers: a responsive frontend hosted on Hugging Face Spaces using the Streamlit SDK, and a robust inference logic layer designed for the AMD Instinct™ MI300X. By utilizing the Pareto Principle, we focused on the core 20% of features—mood-to-metadata mapping—that provide 80% of the user value. The project successfully demonstrates the portability of open-source models like Gemma, proving that developers can achieve enterprise-grade performance and data sovereignty using the AMD Developer Cloud. This end-to-end workflow, from local prototyping in VS Code to public deployment, serves as a blueprint for scalable, cost-effective AI agents.