1. Introduction and basics

• Objectives, procedure, and alignment of expectations
• Value creation through generative AI in companies
• Classification of RAG in modern AI architectures
• Typical RAG use cases and limitations

2. Fundamentals of Retrieval-Augmented Generation

• Functionality and architecture of RAG systems
• Interplay between data, retrieval, and generation
• Common sources of error and quality issues
• Examples and best practices from real projects

3. Chunking and embeddings in practice

• Intuitive understanding of embeddings
• Chunking strategies and their effects
• Visualization of similarities in embedding space
• Hands-on implementation in Python notebooks

4. Retrieval, reranking, and generation

• Similarity Search and Top-K Retrieval
• Reranking strategies for better results
• Prompt design for RAG-based responses
• Implementation of a complete retrieval pipeline

5. Evaluation and optimization

• Why evaluating RAG systems is not trivial
• Quality metrics and automatic evaluation
• Systematic optimization of pipelines
• Comparison, parameter tuning, and traceability

6. Production deployment and MLOps

• Systematic optimization of RAG pipelines
• Parameter search, comparison, and traceability
• Experiment tracking & versioning (e.g., with MLflow)
• Implementation as a service with APIs and monitoring
• Deployment with FastAPI

7. Monitoring and drift

• Why RAG systems deteriorate over time
• Types of drift and their effects
• Practical drift analysis with modified data set
• Derivation of measures