Get in Touch
 Duration 14 hours

Course Outline

Foundations of Retrieval-Augmented Generation (RAG)

  • Defining RAG and its significance in enterprise AI strategies
  • Core components: retrievers, generators, and document storage
  • Distinguishing RAG from standalone LLMs and pure vector search methods

Constructing a RAG Pipeline

  • Setup and configuration of frameworks such as Haystack
  • Processes for document ingestion and data preprocessing
  • Linking retrievers to vector databases like FAISS or Pinecone

Refining the Retriever Component

  • Training dense retrievers with specialized domain data
  • Leveraging sentence transformers and contrastive learning approaches
  • Assessing retriever efficacy using top-k accuracy metrics

Refining the Generator Component

  • Choosing appropriate base models (e.g., BART, T5, FLAN-T5)
  • Differentiating between instruction tuning and supervised fine-tuning
  • Implementing LoRA and other PEFT methods for efficient model updates

Performance Evaluation and Optimization

  • Key metrics for RAG assessment (e.g., BLEU, EM, F1 scores)
  • Managing latency, improving retrieval precision, and minimizing hallucinations
  • Utilizing experiment tracking for continuous iterative improvement

Deployment and Enterprise Integration

  • Integrating RAG into internal search engines and conversational interfaces
  • Addressing security, data access protocols, and governance standards
  • Connecting with APIs, dashboards, and enterprise knowledge portals

Industry Case Studies and Strategic Best Practices

  • Applications in finance, healthcare, and legal sectors
  • Strategies for managing domain drift and updating knowledge bases
  • Emerging trends and future advancements in retrieval-augmented LLMs

Wrap-up and Pathways for Further Development

Requirements

  • Foundational knowledge of natural language processing (NLP) principles
  • Practical experience with transformer-based language architectures
  • Proficiency in Python and an understanding of basic machine learning pipelines

Target Audience

  • Specialists in Natural Language Processing (NLP)
  • Teams responsible for knowledge management

Number of participants


Price per participant

Upcoming Courses

Related Categories