🚢 Titanic ML API

End-to-end ML pipeline: Random Forest (82%) + MLflow tracking → FastAPI REST API → Docker containerized → CI/CD with GitHub Actions

---

Problem

Predict whether a Titanic passenger would survive based on their details — class, age, sex, fare, and family size.

Solution

Trained a Random Forest classifier on 891 passengers with MLflow experiment tracking. Served the best model (82% accuracy) via a FastAPI REST API. Containerized with Docker. Automated testing with pytest and CI with GitHub Actions.

Architecture

Raw Data → Feature Engineering → Model Training (MLflow) → FastAPI /predict → Docker Container → GitHub Actions CI

MLflow Experiments

How to Run

Option 1 — Docker (recommended)

docker pull USERNAME/titanic-api:v1.0 docker run -p 8000:8000 USERNAME/titanic-api:v1.0

Option 2 — Local

git clone https://github.com/icy2701/titanic-ml-api.git cd titanic-ml-api pip install -r requirements.txt uvicorn src.main:app --reload

Open: http://localhost:8000/docs

Run Tests

pytest tests/ -v

MLflow Experiments

cd notebooks mlflow ui --port 5001 Open: http://localhost:5001