Nb. days
4
Face-to-face?
Yes
Remote?
Yes
Min. participants
2
Max. participants
8
Place
Toulouse, Paris, etc.
Pricing
2800€ HT
Artificial Intelligence is experiencing unprecedented acceleration and transformation. While Large Language Models (LLMs) based on the Transformer architecture have revolutionized natural language processing, their enterprise deployment still faces major hurdles: knowledge frozen in time, high operational costs, and the persistent risk of hallucinations. Retrieval-Augmented Generation (RAG) provided an initial, essential solution to these challenges by connecting models to external data sources, thereby ensuring more factual, accurate, and contextualized responses.
However, industry demands are evolving: linear applications are no longer enough. Today, we are witnessing a paradigm shift from purely reactive AI to Agentic AI. Organizations now require autonomous agents capable of planning actions, reasoning in loops, leveraging business tools, and collaborating with one another to achieve complex goals without requiring constant human intervention. This is precisely the purpose of LangGraph, the most powerful orchestration framework for modeling sophisticated reasoning loops and multi-agent systems.
This intensive 4-day training program is built to drive this decisive transition. Designed for technical professionals, it offers a seamless progression from LLM foundations and mastering RAG architectures to the design of autonomous, dynamic agents.
However, industry demands are evolving: linear applications are no longer enough. Today, we are witnessing a paradigm shift from purely reactive AI to Agentic AI. Organizations now require autonomous agents capable of planning actions, reasoning in loops, leveraging business tools, and collaborating with one another to achieve complex goals without requiring constant human intervention. This is precisely the purpose of LangGraph, the most powerful orchestration framework for modeling sophisticated reasoning loops and multi-agent systems.
This intensive 4-day training program is built to drive this decisive transition. Designed for technical professionals, it offers a seamless progression from LLM foundations and mastering RAG architectures to the design of autonomous, dynamic agents.
Upon completion of this curriculum, you will master the entire value chain: you will be able to effectively inject your domain expertise and orchestrate robust, resilient, and collaborative agent graphs to meet your company’s real-world needs.
Goals
- Master the foundations and limitations of LLMs: Understand the theoretical and practical basics of the Transformer architecture, as well as the challenges and limitations of Large Language Models.
- Develop production-grade RAG architectures: Master the key concepts of Retrieval-Augmented Generation and build high-performance pipelines, from data indexing to response generation.
- Grasp the strategic value of Agentic AI: Comprehend the fundamental principles of Agentic AI (planning, reflection, and action cycles) to move beyond the limitations of traditional reactive systems.
- Master the technical components of LangChain and LangGraph: Get hands-on with the fundamental software building blocks (LCEL, chains, graphs, shared states, nodes, and edges) required to model dynamic workflows.
- Design autonomous and resilient agents: Implement application autonomy by building agents capable of leveraging tools and reasoning iteratively via the ReAct framework.
- Orchestrate complex multi-agent systems: Model, develop, and supervise collaboration as well as workflow routing between multiple specialized agents.
To know the dates of the next training sessions,
to adapt this program to your needs
or to obtain additional information,
contact us!
to adapt this program to your needs
or to obtain additional information,
contact us!
Teaching method
- This training is 70% practical and uses illustrated and didactic exercises.
- A daily evaluation of the acquisition of the knowledge of the day before is carried out.
- A synthesis is proposed at the end of the training.
- An evaluation will be proposed to the trainee at the end of the course.
- A course support will be given to each participant including slides on the theory and exercises.
- A sign-in sheet for each half-day of attendance is provided at the end of the course as well as a certificate of completion if the trainee has attended the entire session.
- A follow-up and an exchange with the participants will take place a few days after the training.
Prerequisite
Knowledge of the Python language is necessary. Basic knowledge of machine learning is required.Targets
- Data Engineers
- Data Scientists
- Developers
Evaluation
This training does not require a formal assessment of learningProgram
From LLMs to RAG Architectures - Understanding the Foundations
Large Language Models (LLMs)
- Historical evolution of Artificial Intelligence toward LLMs.
- Fundamental principles of neural networks applied to text processing.
- Concepts of vector representations and embeddings.
- The core of the revolution: the global architecture of Transformer models.
- Landscape and market overview of the most popular models.
- Critical analysis of LLM limitations, introduction to key concepts, and RAG architecture.
Deep Dive into Transformers
- Analysis and breakdown of the foundational research paper "Attention Is All You Need".
- Mathematical and logical understanding of attention mechanisms.
- Mastering structural components: skip connections, tokens, positional encoding, and embeddings.
LangChain and the Industrialization of RAG Pipelines
Getting Started with the LangChain Ecosystem
- Comprehensive overview of the LangChain framework and its full range of capabilities.
- The art of assembly with LangChain Expression Language (LCEL): manipulating Runnables and designing Chains.
RAG Architecture and Engineering
- Ingestion Pipeline: Document chunking strategies, embedding model selection, and vector database utilization.
- Retrieval & Generation Pipeline: Advanced information retrieval algorithms, memory management methods, and context optimization.
Prompt Engineering and Reasoning Strategies
- Implementing prompting techniques: zero-shot, few-shot, Chain-of-Thought, Tree-of-Thought.
- Introduction to the ReAct framework.
The Emergence of Agentic AI and Orchestration with LangGraph
Agentic AI: Theory and Challenges
- Defining Agentic AI through its autonomy cycle: Perception, Reasoning, Action, and Learning.
- Comparative analysis: Traditional agents vs. basic Generative AI vs. proactive Agentic AI.
- The ReAct (Reasoning and Acting) model and its pivotal role in system autonomy.
- Synergy between prompt configuration and external tool integration to ensure reliable behavior.
Core Concepts and Foundations of LangGraph
- Why rely on LangGraph to orchestrate complex agents?
- Core architectural concepts: Graphs, Shared States, Nodes, and Edges.
- Modeling and structuring a graph State for fine-grained control over input/output flows.
Agent Robustness and Reliability
- Advanced management of global state and application memory mechanisms.
- Self-correction patterns and resilience strategies against failures.
Multi-Agent Orchestration, Complex Workflows, and Quality
Designing Multi-Agent Architectures
- Mapping and exploring different multi-agent architectures.
- Analyzing interaction models: sequential execution (Pipeline) vs. open exchange spaces (Group Chat).
- Dynamic routing of communication flows using a Router Agent.
Modeling Complex Systems
- Designing and implementing structured business processes as chained workflows.
- Integrating Human-in-the-loop approval loops and result consolidation mechanisms.
Quality, Observability, and AgentOps
- Ensuring production readiness: tools for observability, flow logging, execution tracing, evaluation, and performance metric tracking.
