The combined scalability and efficiency of Intel® hardware and the flexible power of the IBM watsonx* AI portfolio make an ideal platform for gaining the most from retrieval augmented generation (RAG). This session guides developers to create and launch models that use the advanced capabilities of these hardware and software platforms. It also highlights the advanced optimizations that make Intel® Xeon® processors an ideal choice for hosting embedding models and efficiently handling embedding queries within a RAG pipeline at scale.

Other highlights of the session include a demonstration of the RAG pipeline with an embedding server, a Milvus* vector data store, and an LLM inference with examples illustrating the capabilities of the 4th gen Intel® Xeon® Scalable processor when used with popular Llama models.

The webinar is designed for intermediate-level AI researchers, data scientists, and machine learning engineers who work with RAG pipelines, optimized context-aware LLMs, and vector searches. Technology architects and enterprise developers may also gain new techniques for optimizing AI workloads on Intel hardware.

The following topics are covered in this session:

• Optimizing AI workloads on 4th gen Intel Xeon Scalable processors to capitalize on Intel® Advanced Matrix Extensions (Intel® AMX) and Intel® Extension for PyTorch* to optimize and improve the embedding model inference with AI.
• Designing and deploying a RAG pipeline, which includes embedding generation, vector searches using Milvus, and LLM inference.
• Integrating watsonx for AI applications. Explore techniques for building and scaling context-aware AI solutions with optimized embedding models.
• Maximizing performance of Llama models, which covers enterprise-level techniques for optimizing Llama models on a RAG-based pipeline.
• Gaining hands-on experience with the RAG pipeline and 4th gen Intel Xeon Scalable processors.