Course : Designing and Scaling RAG Architectures with LLMs

Practical course - 2d - 14h00 - Ref. LLR
Price : 1430 € E.T.

Designing and Scaling RAG Architectures with LLMs



New course



INTER
IN-HOUSE
CUSTOM

Practical course in person or remote class
Disponible en anglais, à la demande

Ref. LLR
  2d - 14h00
1430 € E.T.






Teaching objectives
At the end of the training, the participant will be able to:
Identify the key components of a RAG architecture and analyze use cases tailored to business needs.
Implement a RAG pipeline that integrates document ingestion, vectorization, and augmented generation.
Design a secure RAG architecture that takes governance and compliance constraints into account.
Optimize and industrialize a RAG system with hybrid search, reranking, and evaluation mechanisms.

Intended audience
IT project managers, software architects, developers, data engineers, innovation managers, and professionals interested in designing, leading, or scaling projects that incorporate LLMs and RAG architectures.

Prerequisites
Basic knowledge of Python or JavaScript development and a general understanding of web APIs. A basic understanding of artificial intelligence, data, or document processing is recommended.

Course schedule

1
Understanding the Fundamentals of RAG Architectures

  • Identify the limitations of LLMs and the benefits of RAG architectures in business applications.
  • Understanding the stages of a RAG pipeline: ingestion, embeddings, indexing, retrieval, and generation.
  • Analyze the principles of chunking, dense and sparse embeddings, and vector search.
  • Compare the naive, advanced, and agent-based RAG approaches based on use cases.
  • Examine recent trends: Graph RAG, multi-hop retrieval, Self-RAG, and Reflexion.
Hands-on work
Getting started with a simple RAG pipeline and analyzing how its components work.

2
Implement an operational RAG pipeline

  • Set up a document ingestion pipeline for PDF files and Markdown content.
  • Perform document cutting, vectorization, and indexing in a vector database.
  • Configure the interactions between the LLM, the vector search engine, and augmented generation.
  • Analyze the responses and identify the mechanisms that limit hallucinations.
  • Use vector databases such as FAISS, Chroma, Pinecone, or Weaviate.
Hands-on work
Development of a functional RAG on a domain-specific document corpus.

3
Designing a RAG architecture tailored to business needs

  • Formulate a business use case and identify the expectations of users and stakeholders.
  • Evaluate documentary sources, data quality, and volume constraints.
  • Select embedding models, chunking strategies, and architectural components.
  • Design an ingestion pipeline that is consistent with performance and operational constraints.
  • Address compliance, data security, and access management issues.
Hands-on work
Case study and design of a comprehensive RAG architecture that addresses a business need.

4
Securing and Governing a RAG System

  • Formulate a business use case and identify the expectations of users and stakeholders.
  • Implement mechanisms to control access to and secure sensitive data.
  • Apply the principles of data encryption at rest and in transit in a RAG environment.
  • Establish governance rules for the use of LLMs and internal data.
  • Assess regulatory impacts and compliance requirements, including the GDPR.
Hands-on work
Case study and design of a comprehensive RAG architecture that addresses a business need.

5
Optimizing Retrieval Performance and Quality

  • Implement hybrid search mechanisms that combine BM25 and vector search.
  • Use reranking techniques with cross-encoders to improve the relevance of responses.
  • Use metadata filters and multi-user management in a professional setting.
  • Measure performance using Precision@k, Recall, and MRR.
  • Identify ways to improve quality and response times.
Hands-on work
Optimization of a RAG pipeline using hybrid search and evaluation of results.

6
Industrializing and Deploying a Business Assistant Based on RAG

  • Scale up an ingestion pipeline with advanced parsing, data cleansing, and data enrichment.
  • Implement mechanisms for versioning and data lifecycle management.
  • Understand the principles of agents and tool-augmented RAG to enhance interactions.
  • Prepare for the deployment and operation of an RAG system in a professional environment.
Hands-on work
Development of a business information assistant based on an advanced RAG architecture.



This programme is an original creation, developed by the teaching teams at ORSYS Formation. Any reproduction, representation, adaptation or use, in whole or in part, without the prior written authorisation of ORSYS, is strictly prohibited. ORSYS reserves the right to take any action necessary to protect its intellectual property rights.

Dates and locations
Select your location or opt for the remote class then choose your date.
Remote class

Dernières places
Date garantie en présentiel ou à distance
Session garantie

REMOTE CLASS
2026 : 10 Dec.

2027 : 22 Feb., 3 May, 8 July, 11 Oct.

PARIS LA DÉFENSE
2026 : 3 Dec.

2027 : 22 Feb., 3 May, 8 July, 11 Oct.



This programme is an original creation, developed by the teaching teams at ORSYS Formation. Any reproduction, representation, adaptation or use, in whole or in part, without the prior written authorisation of ORSYS, is strictly prohibited. ORSYS reserves the right to take any action necessary to protect its intellectual property rights.