RAG ATTACK: Step into the Era of Agentic RAGs

GENERATIVE AI
An advanced training program co-designed with Sorbonne University: learn how AI agents can orchestrate complex RAG pipelines, and explore the tools and cloud platforms to deploy your solutions at scale.

Objectives

• Understand the foundations and limitations of RAG (Retrieval-Augmented Generation) systems.
• Discover the concept of agentic RAGs, at the intersection of retrieval, generation, and autonomous planning.
• Implement a RAG pipeline powered by agents.
• Identify real-world use cases and explore the tools to build and deploy these architectures.

Program

MORNING – RAG Systems (Led by One Point)

9:00–9:15 AM: Welcome coffee

9:15–10:45 AM: Lecture – Fundamentals of RAG

Theoretical content:

• Context and historical evolution of RAG
• Technical fundamentals: embeddings, prompts, LLMs
• Architecture of a RAG pipeline: chunking, vector database, retrieval
• Performance optimization and current limitations
  15-minute break

11:00 AM–12:30 PM: Hands-on Workshop – Building a RAG System
Progressive practical exercises:

1.Exploring VéloCorp data

• Exploration of the document corpus (catalogs, manuals, policies)
• Structure analysis and identification of use cases

2. Building a local vector database (FAISS)

• Chunking: intelligent document segmentation and experimentation with chunk sizes
• Embedding: vector transformation, model comparison
• Indexing and local persistence

3. Optimizing Retrieval

• Cosine similarity search techniques
• Prompt impact: simple vs enriched queries
• Ranking and re-ranking strategies

4. Scaling up with Azure Cognitive Search

• Migration to a professional cloud solution
• Index and analyzer configuration

12:30–1:30 PM: Lunch break (free)

 
AFTERNOON – Agentic AI (Led by SCAI)

1:30–3:00 PM: Lecture – Fundamentals of Agentic AI
Theoretical content:

• Agent context: transformation brought by LLM-based agents and main components
• Main agentic architectures
• Deployment and security
• Future perspectives and evolutions
  15-minute break

3:15–4:45 PM: Hands-on Workshop – Building Intelligent Agents
Practical exercises with increasing levels of complexity:

Level 1 – Simple Agent
Automatically querying the appropriate source based on the question

Level 2 – ReAct Agent
Reasoning before acting

Level 3 – Multi-tool Orchestration
Combining multiple sources for complex analyses

Level 4 – Hierarchical Planning
Decomposing and executing complex tasks

4:45–5:00 PM: Wrap-up and Outlook

• Feedback session
• Concrete applications in your professional projects
• Q&A

Target Audience
AI engineers, data scientists, and developers.

Prerequisites
Basic knowledge of Python, APIs, and large language models (LLMs).

Assessment Methods
Hands-on exercises and case studies are integrated throughout the course to validate learners' understanding.

Teaching Methods and Resources
This concept-driven course combines instructional content with practical exercises, workshops, and case studies for an applied learning experience.

 

To apply :

• https://fc.sorbonne-universite.fr/nos-offres/rag-attack-entrez-dans-lere-des-rag-agentiques/
• https://ecole.groupeonepoint.com/product/formation-intelligence-artificielle-rag-attack-entrez-dans-l-ere-des-rag-agentiques