The University of Queensland

Advanced Techniques for High Dimensional Data (INFS4205/7205)

This course explores how modern AI systems represent, align, retrieve, and reason over high-dimensional multimodal data. Contemporary AI systems, including multimodal foundation models, retrieval-augmented generation (RAG), and agents, rely on representing text, images, and other modalities as high-dimensional embeddings. Understanding how these representations are learned, aligned, indexed, and used for reasoning is essential for building reliable AI applications.

The course takes a systems view of multimodal AI: from representation learning and alignment to vector databases and scalable multimodal RAG, followed by efficiency techniques and agent architectures that enable interactive AI systems. It concludes with frontier applications and personalization methods such as parameter-efficient fine-tuning. A central theme is vibe coding: designing AI-assisted systems by reasoning about intent, structure, and evaluation rather than syntax alone. Through reflective inquiry and a hands-on multimodal chatbot project, students learn to build, analyze, and critically evaluate modern AI pipelines.