A data science portfolio people actually read

Why a repo full of notebooks is not enough

Most people who look at a data science portfolio will not run your notebook. Hiring managers skim, and many recruiters are not technical. What they look for is whether you can pick a sensible question, handle messy data and explain a result to someone who has to act on it.

So write that part down. On makerhandle every project gets a project page with a Markdown write-up, screenshots of your charts with captions, the tools you used as tags, your role and timeline, and highlights such as "cut forecast error by 18% vs the baseline".

A structure that works

For each project, a few short paragraphs:

1. The question and why it mattered, in plain words.
2. The data: where it came from, how big, what was wrong with it.
3. What you tried, including what did not work.
4. The result, against a baseline, with one chart.
5. What you would do next.

Link the source (GitHub or Kaggle), a live demo (a Streamlit app or a Hugging Face Space), and a short video walkthrough. Our guide to writing a project case study goes deeper.

Pick three good projects

Three to five projects that each show something different beat ten tutorials. Avoid the classic datasets everyone has seen unless you did something new with them, and say what you added. A project on data you collected yourself, or one that answers a question from a real job or hobby, is far more memorable. See what counts as a portfolio project.

Open to work, resume and your own domain

Turn on Open to work for a badge by your name and a contact button on every project page, and upload your resume as a PDF. Analytics show which projects get opened. Free covers five projects; Pro is $20 a year for unlimited projects and a custom domain. The portfolio link for your resume page covers the job search side.