Teaching

Teaching materials

📚 Dauphine Tunis 2025-2026 - Generative AI

📋 Project Instructions:

Check this code repository: https://github.com/BastinFlorian/dauphine-project-iasd-2025

📅 Today’s Course (November 4, 2025)

🎓 Introduction to LangGraph

⏰ Deadlines

📖 Deadline 1: November 13, 2025 - 23:59

Learning Materials & LangGraph Course

  • What to complete:
    • All resources from the “Mandatory Learning Path” section
    • LangChain Academy - Introduction to LangGraph course
  • ⚠️ No submission required but an exam may be held during the class on November 14, 2025

💻 Deadline 2: November 13, 2025 - 23:59

Google Colab Notebooks

  • What to submit:
    • All completed Google Colab notebooks from the course
  • How to submit:
    • Send by email a ZIP file containing all completed notebooks or send me a mail with a sharing access to a drive repository
  • ⚠️ Important: No submissions will be accepted after this deadline

🎯 Deadline 3: December 15-16, 2025

Final Project Presentations

  • 📋 Project topic: Check this link
  • Presentation dates:
    • December 15, 2025: 12:15 - 13:45
    • December 16, 2025: 12:15 - 13:45
  • 📝 Registration: Sign up for your presentation slot here
  • What to prepare:
    • Final project demonstration
    • Project code and documentation

📚 Course Materials by Chapter

The courses slides have been sent by email with the following object “Generative AI - Slides de cours”.

📖 Mandatory Learning Path (Follow in Order)

# Resource Author/Source Study Time Link Description
1 Visual Understanding of LLMs 3Blue1Brown 2 hours YouTube Playlist Beautiful visual explanations of neural networks and deep learning fundamentals
2 Building Large Language Models Stanford CS229 1.5 hours YouTube Stanford’s comprehensive lecture on building LLMs from scratch
3 How Transformer LLMs Work DeepLearning.AI 2 hours Course Comprehensive course on Transformer architecture and LLM fundamentals
4 Andrej Karpathy’s Neural Network Lectures Andrej Karpathy 2 hours YouTube Building GPT from scratch, excellent for understanding fundamentals
5 The Illustrated Transformer Jay Alammar 1 hour Blog The best visual guide to understanding Transformers, step-by-step
6 Attention Mechanism Visualized Jay Alammar 1 hour Blog Visual breakdown of attention mechanisms in sequence models
7 Attention Is All You Need Vaswani et al. 3 hours arXiv The foundational paper introducing Transformers, basis of all modern LLMs
8 BERT: Pre-training of Deep Bidirectional Transformers Devlin et al. 2 hours arXiv First massively pre-trained model that revolutionized NLP, introduced transfer learning
9 The Illustrated GPT-2 Jay Alammar 30 mins Blog Visual explanation of GPT architecture and training
10 DeepSeek-R1: Incentivizing Reasoning Capability in LLMs DeepSeek-AI 2 hours arXiv Breakthrough in open-source reasoning models, matching OpenAI o1 performance

Total Study Time: ~16.5 hours

💻 Code Repository

📚 Notebooks Google Colab

# Title Key topics Lien Google Colab
LANGCHAIN FUNDAMENTALS      
1 First LLM Call API setup, ChatGoogleGenerativeAI, invoke(), temperature, max_tokens, model comparison Open In Colab
2 Prompt Templates PromptTemplate, ChatPromptTemplate, input variables, template formatting, system/human messages Open In Colab
3 Message Types HumanMessage, AIMessage, SystemMessage, message roles, conversation structure Open In Colab
4 Structured Outputs Pydantic models, BaseModel, type validation, structured data extraction, JSON parsing Open In Colab
5 Simple Chains LCEL (pipe operator), chain composition, RunnablePassthrough, data flow Open In Colab
6 Multimodal Inputs Image processing, base64 encoding, vision models, multimodal prompts Open In Colab
7 Streaming Responses Real-time streaming, AIMessageChunk, chunk aggregation, streaming chains Open In Colab
LANGGRAPH FUNDAMENTALS      
1 LangGraph Basics StateGraph, TypedDict, nodes, edges, conditional routing, state persistence Open In Colab
2 Tool Calling Tool definitions, function calling, StructuredTool, tool integration, parameter validation Open In Colab
3 ReAct Prebuilt Agent ReAct pattern, create_react_agent, agent executor, reasoning and acting, tool selection Open In Colab
4 Langfuse Callback Handler Observability, tracing, callback handlers, monitoring, performance tracking Open In Colab

📚 Dauphine Tunis 2024-2025 (Archive)

📖 Course Materials

☁️ AI with Google Cloud Platform

🎯 Final Project & Grading (2024-2025)

📋 Project Overview: Create a technical chatbot for answering medical field questions using cloud technologies.

📊 Available Dataset:

✅ Core Requirements (1-7)

# Component Description
1 🖥️ Streamlit Interface Interactive web interface for question-answering
2 ☁️ Cloud SQL Setup Database integration with Streamlit/Gradio interface
3 💬 Exact Answers Precise answer generation for each question
4 📚 Source Display Show sources with similarity scores and focus areas
5 💻 Code Quality GitHub repo with README, requirements, and best practices
6 🚀 Deployment Cloud Run deployment with project naming convention
7 📈 Evaluation Metrics selection and eval.py script for testing

🏆 Additional Requirements

  • 📋 Model Card: Comprehensive documentation of your solution
  • 🎨 Bonus Creativity: LangGraph agents, LangFuse monitoring, feedback dashboards, fine-tuning, multimodality

📅 Submission Details

  • 📆 Deadline: January 15th (GitHub link + model card)
  • 🎤 Presentation: January 25th (15 min demo + Q&A)

🤖 Generative AI Course

  • 📝 Lab Reports + Lab 4: 20%
  • 🧠 Quiz Results: 30%
  • 🎯 Final Project: 50%

☁️ AI on the Cloud Course

  • 📝 Lab Reports + Lab 4: 40%
  • 🎯 Final Project: 60%

🎯 Evaluation Criteria

For Generative AI:

  • ✨ Quality of reasoning and model choices
  • 🚀 Performance of your solution
  • 🔬 Rigor of evaluation
  • 🎤 Quality of presentation

For AI on the Cloud:

  • ☁️ Thoughtful use of Cloud tools
  • 📁 Professional GitHub repository
  • 📏 PEP8 compliance (Flake8, PyLint, Black)
  • 🔄 Good Git practices (commits, PRs)