🧠 Lexo - LLM Toolkit for RAG, Fine-tuning, and AI Agents

Lexo is a comprehensive collection of Jupyter notebooks designed for learning and applying Large Language Models (LLMs) in real-world scenarios. Master RAG systems, fine-tuning (QLora), AI agents, multimodal processing, and ML benchmarks through hands-on projects.

🌐 Visit our website: https://lexo.lisekarimi.com

This notebook is inspired by inspired by Ed Donner's LLM engineering certification.

• Udemy Course: https://www.udemy.com/course/llm-engineering-master-ai-and-large-language-models/
• GitHub Repository: https://github.com/ed-donner/llm_engineering

🛠️ Prerequisites & Installation

• Python 3.11.x (not 3.12+)
• uv package manager
• Make: winget install GnuWin32.Make (Windows) | brew install make (macOS) | sudo apt install make (Linux)
• Any IDE of your choice (e.g., Visual Studio, Cursor, PyCharm).

🚀 Quick Start

# Clone the repository
git clone https://github.com/lisekarimi/lexo.git
cd lexo

# Setup environment
mv .env.example .env  # Add your API keys here

# Install dependencies
uv sync

Select Jupyter Kernel:

1. Open any .ipynb file
2. Click "Select Kernel" in the top-right corner of your IDE
3. Choose "Python Environments" → Select the .venv interpreter

🔧 Projects & Notebooks

Core AI Applications

• 🌐 Website Summarizer – Summarize any URL using OpenAI + LLaMA
• 🧾 Brochure Generator – Transform websites into AI-crafted brochures
• 💡 Tech Assistant – Explain complex tech concepts in simple terms
• 🤖 AI Chat – Multi-model conversations with GPT, Claude & Deepseek
• 🌤️ Weather AI – Location-based activity suggestions using weather data

Advanced Workflows

• 📝 Meeting Minutes Assistant – Audio transcription + structured notes (Whisper + LLaMA)
• 🧪 Synthetic Data Generator – Create custom datasets for ML projects
• 🧠 Internal RAG Assistant – Enterprise knowledge base with fast retrieval

ML & Fine-tuning Pipeline

• 💲 Price Predictor – Complete ML pipeline from data to deployment:
  • Data curation and preprocessing
  • Traditional ML vs. LLM benchmarking
  • RAG experiments with embeddings
  • Fine-tuning frontier models + QLoRA for open-source models
  • Comprehensive evaluation and leaderboard

Capstone Project

• 🏷️ Snapr – Multi-agent AI system for finding the best deals

🎯 Key Features

Core Skills Covered:

• Retrieval-Augmented Generation (RAG)
• Prompt engineering and optimization
• AI agents and tool integration
• Fine-tuning and QLoRA techniques

Models & Technologies:

• Frontier LLMs (OpenAI GPT, Anthropic Claude, ...)
• Open-source models (LLaMA, Mistral, Deepseek, ..)
• Speech-to-Text processing
• Multimodal input handling
• Vector databases and embeddings

📚 Learning Path

Each notebook includes:

• Clear problem statement and real-world context
• Step-by-step implementation with explanations
• Required API keys and setup instructions
• Customizable parameters for experimentation
• Performance evaluation and insights

🔗 Related Projects

📸 Snapr

• Demo
• Code

📊 DataGen

• Demo
• Code

🎯 Getting Started

1. Choose your setup method (Docker recommended for beginners)
2. Configure your API keys in .env file
3. Start with a simple project like Website Summarizer
4. Progress through notebooks based on your interests
5. Experiment and customize the code for your use cases

💡 Tips

• WebPage Summarizer notebook requires Chrome - run this locally if using Docker
• Start with projects matching your experience level
• Each notebook is self-contained - jump around as needed

---

Ready to dive into the world of LLMs? Choose your setup method and start exploring! 🚀