roadmaps / data-scientist
How to become a data scientist
A practical, stage-by-stage path from beginner to hireable data scientist — the math, the tools, and the order to learn them in, with free resources at every step.
› Work the stages in order — the math and Python foundations make everything after them far easier.
› Every resource here is free. Mark a stage passed when you've built something with it, not just watched a video.
› The portfolio matters more than any certificate. Employers hire proof you can turn data into decisions.
Data science sits at the intersection of statistics, programming, and domain knowledge — you turn messy data into insight and predictions. It's one of the best-paid fields in tech, and everything you need to learn it is free. The trap is jumping straight to fancy machine learning; the people who actually get hired build a solid foundation first, then prove it with projects. These eight stages take you from "what's a mean?" to a portfolio that gets interviews.
STAGE 01 / 08
Math & statistics
Statistics is the language of data science — it's how you tell signal from noise and quantify uncertainty. Don't try to memorize formulas; build intuition first. Learn descriptive stats, probability, distributions, sampling, hypothesis testing, and the idea behind regression. A little linear algebra and calculus help later, but stats is the priority.
The good news: the best resources make this genuinely enjoyable. Get the visual, plain-language explanations down before you touch a single model.
- roadmap.sh — AI & Data Scientist ↗An interactive visual map of data-science skills — a great big-picture companion to this page.roadmap.sh
- StatQuest ↗Josh Starmer's free videos break down statistics and ML with unbeatable clarity. The single best intuition-builder.statquest.org
- → Our Statistics resource listStatQuest, Seeing Theory, Khan Academy, and Think Stats — intuition first, then rigor.studylistsvault.com
STAGE 02 / 08
Python for data
Python is the language of data science — readable, ubiquitous, and backed by the best data ecosystem anywhere. Learn the fundamentals (data types, loops, functions, working with files) and get comfortable in Jupyter notebooks, where data scientists live day to day.
You don't need to become a software engineer — you need to be fluent enough to load data, transform it, and express an analysis clearly. Learn by doing, on real datasets.
STAGE 03 / 08
Data wrangling: pandas & SQL
Real data is messy, and you'll spend most of your time cleaning and reshaping it. pandas is the tool for that — filtering, grouping, joining, and reshaping tables in Python. Master it and you can wrangle almost any dataset into shape.
Equally essential is SQL. Most data lives in databases, and every data role assumes you can query it. Learn SELECTs, JOINs, GROUP BY, and window functions — it's the most under-rated data-science skill.
- pandas Documentation (official) ↗The workhorse of every data-science project, with a great "10 minutes to pandas" intro.pandas.pydata.org
- SQLBolt ↗Free, interactive, in-browser SQL lessons — the best place to learn SQL by doing.sqlbolt.com
- → Our SQL for Data Science listSQLBolt, Mode's tutorial, and Kaggle Learn.studylistsvault.com
STAGE 04 / 08
Data visualization
A finding nobody understands changes nothing. Visualization is where data science meets communication — turning numbers into clear, honest charts that drive decisions. Learn Matplotlib and Seaborn for exploratory and statistical plots, and develop judgment about which chart fits which question.
This skill compounds: better charts during exploration help you find insights faster, and better charts in a report get your work acted on.
STAGE 05 / 08
Machine learning
Now the payoff: teaching computers to find patterns and make predictions. Start with the classics — linear and logistic regression, decision trees, random forests, and clustering — and the workflow around them: train/test splits, cross-validation, and avoiding overfitting. Andrew Ng's course is the legendary on-ramp; scikit-learn is the library you'll build with.
Focus on understanding when and why to use each model, not just how to call it. That judgment is what separates a data scientist from someone who copies notebooks.
STAGE 06 / 08
Deep learning
Deep learning powers modern AI — image recognition, language models, and more. You don't need it for every job, but it's increasingly expected and it's fascinating. Learn how neural networks work, then get hands-on with a framework. Andrew Ng's Deep Learning Specialization builds the theory; fast.ai gets you training real models fast.
Take it as far as your goals require — a broad understanding plus one solid project is plenty to start.
STAGE 07 / 08
Projects & portfolio
This is the stage that gets you hired. Courses prove you can follow along; projects prove you can do the job. Build 2–4 end-to-end projects on datasets you find interesting — frame a question, clean the data, analyze it, model it, and communicate the result clearly. Put them on GitHub with a readable write-up.
Kaggle is the perfect playground: real datasets, competitions, and a community. Even a few notebooks that tell a clear story make a strong portfolio.
STAGE 08 / 08
Specialize & land the job
Data science is broad, so pick a direction as you go deeper: natural language processing, computer vision, analytics/BI, or data engineering. A focus makes you memorable and gives your portfolio a theme.
Then prepare for the hunt. Data-science interviews test SQL, statistics, ML concepts, and a case study or take-home. Practice explaining your projects out loud — being able to tell the story of a project is half the battle. You don't need to know everything; you need to prove you can turn data into decisions.