Jupyter Notebook Guide [2026]: Data Science Workflow

Jupyter Notebooks are where data science actually happens. Exploration, EDA, prototyping, and communicating findings all happen in notebooks before code moves to production pipelines. This guide covers what you need to be productive.

Setup and Launch

# Install and launch
pip install jupyterlab ipykernel

# Register a virtual environment as a kernel
python -m ipykernel install --user --name myproject --display-name "My Project"

# Launch JupyterLab
jupyter lab  # opens at http://localhost:8888

Create a separate virtual environment for each data project. Register it as a named kernel so you can switch environments from the JupyterLab kernel selector without leaving the browser.

Keyboard Shortcuts

Running cells:

• Shift+Enter: Run cell and move to next
• Ctrl+Enter: Run cell in place
• Alt+Enter: Run cell and insert new cell below

Command mode (press ESC first):

• A: Insert cell above | B: Insert cell below
• DD: Delete cell | Z: Undo delete
• M: Convert to Markdown | Y: Convert to code
• Shift+L: Toggle line numbers

Edit mode (press Enter):

• Tab: Autocomplete | Shift+Tab: Show docstring
• Ctrl+/: Toggle comment

Magic Commands

# Time a cell
%%time

# Time a single expression
%timeit df.groupby('category').sum()

# Shell commands
!pip install new-package
!ls -la data/

# Auto-reload imported modules (critical for development)
%load_ext autoreload
%autoreload 2

# Show plots inline
%matplotlib inline

Data Science Workflow Structure

Structure every notebook with these sections:

1. Imports and configuration (all imports in cell 1)
2. Data loading and inspection (shape, dtypes, head, describe)
3. Data cleaning (nulls, types, duplicates)
4. Exploratory Data Analysis (distributions, correlations, outliers)
5. Feature engineering (new features, encoding)
6. Modeling (train/test split, model, evaluation)
7. Conclusions (Markdown summary of key findings)

Use Markdown cells between code sections to explain what you are doing and why. A notebook that is just code is documentation for your future self that is already half-forgotten.

Frequently Asked Questions

What is the difference between Jupyter Notebook and JupyterLab?

JupyterLab is the modern successor to the classic Jupyter Notebook interface. JupyterLab adds a file browser, multiple notebook tabs, split view, a text editor, a terminal, and an extension ecosystem. New users should start with JupyterLab.

How do I share a Jupyter Notebook?

Export as HTML (File > Export > HTML) for a static document. Share the .ipynb file directly for others with Jupyter. Use nbconvert for PDF or Markdown export. Host on GitHub (notebooks render automatically). Use Google Colab for cloud-hosted collaborative notebooks.

How do I use Jupyter with Git?

Jupyter notebooks store outputs (plots, print statements) in the .ipynb JSON, creating noisy diffs. Use nbstripout as a pre-commit hook to strip outputs before committing, or use Jupytext to save notebooks as Python files with cell markers that diff cleanly.

What is Google Colab?

Google Colab is a free cloud-hosted Jupyter environment with free GPU/TPU access, pre-installed ML libraries, and no local setup required. It is the best option for learning ML on a machine without a dedicated GPU. Colab Pro ($10/month) provides longer runtimes and more powerful GPUs.

Note: Information reflects early 2026.