Learning Docker

Jackson Hoffart

Learning Docker

• 🚀 Docker makes our code portable & consistent across environments.
• 🔄 We aim to “build once & run anywhere” — local, cloud, CI/CD, etc.
• 🏗️ Docker containers help solve dependency conflicts and simplify collaboration.

The high level pitch. The slide goal is to motivate why Docker matters for data scientists: portability, consistency, and reducing “dependency hell.”

“But, it works on my machine”

• ❓ Has anyone ever had code that runs perfectly locally but fails elsewhere?
• ❗ Reproducibility is hard across machines, OS versions, or teammates
• ✅ Docker helps us create self-contained and portable environments that work anywhere¹

¹ Sort of.

Audience should connect emotionally to the problem — emphasize pain points they’ve likely experienced. The idea is to set Docker up as the solution to a shared frustration.

So, what is Docker?

• A containerization platform
• Bundles code, libraries, and environment into one unit
• Think of it as a “lightweight virtual computer”
• Ensures consistency and reproducibility

Introduce Docker at a high level. Key takeaway: Docker packages everything your Python script needs to run — dependencies, system libraries, etc.

Core Concepts

• Image: The blueprint (like a saved environment)
• Container: A running instance of an image
• Dockerfile: Instructions for how to build an image
• Registry: Where images are stored/shared (e.g. Docker Hub, GHCR)

Audience gets a first look at key terms. You can use analogies like: image = cake recipe, container = actual cake, Dockerfile = recipe card.

Dockerfile: The Recipe

FROM python:3.11-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY . .
CMD ["python", "script.py"]

Goal: Show that defining a Docker image is just writing a simple file. Emphasize how readable and reproducible this is.

Build & Run Your Container

docker build -t my-python-app .
docker run my-python-app

• build: creates an image from a Dockerfile
• run: launches the container using that image

Demystify the commands. Help them remember: Build = prepare environment, Run = execute your code inside it.

Common Data Science Use Cases

• Share Jupyter notebooks with all dependencies
• Run scheduled pipelines reliably
• Train models in consistent environments
• Ensure old analyses still work later
• Integrate easily with CI/CD pipelines

Make Docker real for them. These are specific use cases they’ll care about. Invite them to imagine using Docker in their own projects.

Pro tips: Docker Best Practices

• 📌 Pin all dependency versions
• 📁 Use .dockerignore
• 🧱 Layer instructions efficiently
• 🏷️ Tag images clearly
• 🔒 Avoid root unless absolutely necessary

Expose the “pro tips” that help them avoid common pitfalls as they start using Docker. These tips build good habits early.

Summary

• ✅ Docker solves “it works on my machine”
• ✅ Simple, readable files to define environments
• ✅ Great fit for Python data science workflows
• ✅ Paves the way for production & collaboration

Reinforce the core takeaways before Q&A. Remind them: Docker is simple to start with, and pays dividends quickly.

Q&A

• 💬 Questions, use cases, confusions?

Allow time for discussion or a quick group demo. Optionally offer to live-debug a student’s Dockerfile.

Resources

• Docker tutorial for beginners: https://docker-curriculum.com/
• Docker documentation: https://docs.docker.com/