Kaggle Machine Learning Solutions

This repository contains my solutions to various Kaggle competitions, showcasing machine learning best practices with modern MLOps tooling.

📊 Solutions

Titanic - Machine Learning from Disaster

A comprehensive ML pipeline for the classic Titanic survival prediction challenge, featuring:

• Advanced MLOps Stack: Optuna hyperparameter optimization + MLflow experiment tracking
• Multiple Model Types: Neural Networks, Random Forest, XGBoost, LightGBM
• Production-Ready Code: Modular architecture with proper preprocessing, cross-validation, and model evaluation
• NixOS Environment: Reproducible development environment with GPU support

→ View Titanic Project Details

🚀 Getting Started

Each solution directory contains its own README with specific instructions. Generally:

# Clone the repository
git clone <repository-url>
cd kaggle_solutions

# Enter NixOS development environment (if using Nix)
direnv allow

# Or install dependencies with uv
uv sync

# Navigate to specific solution
cd titanic/

🛠️ Tech Stack

• Python 3.11+
• Package Management: uv
• ML Frameworks: TensorFlow, PyTorch, scikit-learn, XGBoost, LightGBM
• MLOps: Optuna, MLflow
• Environment: NixOS with direnv

📁 Repository Structure

.
├── titanic/              # Titanic survival prediction
│   ├── README.md         # Project documentation
│   ├── data/             # Kaggle datasets
│   ├── models/           # Saved models and parameters
│   ├── titanic/          # Source code
│   └── ...
├── plans/                # Architecture & design documents
├── pyproject.toml        # Root dependencies
└── README.md             # This file

📝 License

This project is for educational and portfolio purposes.

🔗 Links

• Kaggle Profile
• Titanic Competition