AI agents, RAG and MCP
Learning objectives
By the end of the course, participants will be able to:
- Explain how a language model, a RAG system and an agent work
- Build a RAG pipeline: chunking, embeddings, vector database, retrieval and source citation
- Design an agent that calls tools and chains several steps
- Expose and use tools through the MCP protocol
- Evaluate the quality, reliability and cost of a system
Audience
Students in computer science, data science or AI (master’s, engineering school).
Prerequisites
Programming in Python; basic knowledge of web APIs. Prior use of a language model is a plus.
Day-by-day programme
| Session | Content |
|---|---|
| Day 1 morning | Language models: how they work, APIs, system prompts, structured outputs |
| Day 1 afternoon | RAG: chunking, embeddings, vector database, hybrid search, source citation |
| Day 2 morning | Evaluating a RAG system: question sets, relevance and faithfulness measures, common errors |
| Day 2 afternoon | Agents: tool calling, execution loop, memory, guardrails and human validation |
| Day 3 morning | MCP: protocol principles, writing an MCP server, connecting it to a client |
| Day 3 afternoon | Project: document assistant with an agent and MCP tools, demonstration and review |

