LLM Engineering Journey

This repository contains my journey through LLM engineering, inspired by Ed Donner's LLM Engineering Course. The repository is organized into several key components, each representing different aspects of my learning and implementation of LLM technologies.

Repository Structure

1. Pricer Models (/pricer)

A comprehensive collection of price prediction models, ranging from frontier models to open-source implementations:

• Frontier Models (frontier-pricers.ipynb): Implementation of state-of-the-art price prediction models
• Open Source Models (os-pricers.ipynb): Open-source alternatives for price prediction
• Fine-tuned Models:
  • GPT-based fine-tuning (gpt-fine-tuned-pricers.ipynb)
  • Open-source fine-tuning (os-fine-tuned-pricers.ipynb)
• Classic Models (classic-pricers.ipynb): Traditional machine learning approaches
• Data Processing:
  • Data curation (data_curator.ipynb)
  • Data loaders (loaders.py)
  • Item definitions (items.py)

2. Production Project (/project)

A production-ready framework featuring collaborative AI agents:

Agents Framework

• Planning Agent (planning_agent.py): Orchestrates and coordinates other agents
• Messaging Agent (messaging_agent.py): Handles communication between agents
• Scanner Agent (scanner_agent.py): Specializes in data scanning and analysis
• Specialist Agent (specialist_agent.py): Provides domain-specific expertise
• ML-based Agents:
  • Random Forest Agent (random_forest_agent.py)
  • Ensemble Agent (ensemble_agent.py)
• Frontier Agent (frontier_agent.py): Integrates with cutting-edge LLM models

Core Components

• Framework (deal_agent_framework.py): Custom framework for agent collaboration
• Main Application (main.py, main_final.py): Production entry points
• Infrastructure (infra.ipynb): System architecture and deployment details
• Vector Store (products_vectorstore/): Storage for product embeddings

3. Learning Journey (Notebooks)

A collection of Jupyter notebooks documenting my learning process:

• Gradio Applications:

  • Multi-AI Translation System (gradio_multi_ai_translate.ipynb)
  • Flight Information System (gradio_multi_ai_flights.ipynb)
  • Tool Integration (gradio_with_tools.ipynb)
  • Basic Bot Implementation (gradio_bot.ipynb)

• LLM Fundamentals:

  • Personal RAG Implementation (Personal_RAG.ipynb)
  • Transformer Interactions (talk_to_transformers.ipynb)
  • Tokenizer Studies (tokenizers.ipynb)
  • HuggingFace Pipelines (hf_pipes.ipynb)

• Bot Development:

  • Multi-Agent Conversations (talking_bots.ipynb)
  • Ollama Integration (ollama.ipynb)

Inspiration

This repository is heavily inspired by Ed Donner's LLM Engineering Course. The course provided the foundation for understanding and implementing various LLM technologies, from basic concepts to advanced production systems.

Getting Started

Each component of this repository can be explored independently:

1. For price prediction models, start with the /pricer directory
2. For the production agent system, explore the /project directory
3. For learning materials, browse through the various Jupyter notebooks

Requirements

• Python 3.8+
• Jupyter Notebook
• Various ML and LLM libraries (requirements vary by component)

License

This project is licensed under the MIT License - see the LICENSE file for details.