Docker for MLOps¶

Introduction¶

Learn how Docker helps package model-serving code, dependencies, and model artifacts into reproducible inference containers.

Before You Start¶

You need a serving application, dependency file, and a decision about how the model artifact enters the image. Small teams often bake the approved model into the image; larger teams may download approved versions at startup.

Project Structure¶

ml-api/
├── app.py
├── requirements.txt
├── models/churn.pkl
└── Dockerfile

Step-by-Step Deployment¶

Dockerfile example:

FROM python:3.11-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY app.py .
COPY models/ ./models/
EXPOSE 8000
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "8000"]

Build and run:

docker build -t ml-api:churn-17 .
docker run --rm -p 8000:8000 ml-api:churn-17

Successfully tagged ml-api:churn-17
Uvicorn running on http://0.0.0.0:8000

Testing the Deployment¶

Test the container:

curl -s http://localhost:8000/health

{"status":"ok","model_version":"churn:17"}

Production Considerations¶

Pin dependency versions, scan images, set non-root users, keep images small, publish immutable tags, and include model version metadata.

Common Mistakes¶

• Using latest for production deployments.
• Building images from unpinned dependencies.
• Copying training data into the serving image.
• Running containers as root unnecessarily.

Related Guides¶

• Deploy an ML Model with FastAPI
• Kubernetes for MLOps
• CI/CD Pipeline for Machine Learning Projects

Summary¶

A reliable model deployment is versioned, testable, observable, and reversible.