New: Roadmaps ordered paths through our cheat sheets and flashcards, so you always know what to study next.
Explore themSee what's new on GitHubFrom your first Python script to conclusions a business will actually act on.
A 15-step learning path. Follow it in order, or jump to what you need.
This path is for anyone who wants to turn messy, real data into decisions a business can act on, whether you're coming from a general tech background or pivoting out of an analyst role. Plan on about 8 to 11 weeks at a few hours a week, moving from Python and SQL through statistics, experimentation, and a first machine learning model. It does not cover BI dashboard tools like Power BI or Tableau; that ground belongs to the Data Analyst path instead. By the end you can clean a genuinely dirty dataset, design and interpret an A/B test, and turn a model's results into a story a non-technical stakeholder will act on.
No prior experience needed. Start from zero.
Open this first: you can write working Python today, and it's the one skill every later step in this path assumes, from your first Git commit to your last model.
Track the code you just wrote in step 1 with real version control, so your work survives revision instead of living in one fragile untracked folder.
Most real datasets live inside a database, not a spreadsheet, so pair the Python from step 1 with the query skill that actually pulls the rows you'll analyze next.
You can write Python, commit it to Git, and pull real rows out of a database with SQL. Next up: turning that raw pull into something worth analyzing.
Finish this section to unlock.
+100 XP
Turn the rows SQL hands you in step 3 into a workable table: pandas is where that raw query result gets cleaned, reshaped, and made ready for analysis.
See what actually runs underneath pandas in step 4: vectorized arrays are why your cleaning and math stay fast instead of crawling row by row.
Put pandas and NumPy to work on a genuinely messy dataset: outliers, missing values, and distributions, the pass that decides whether anything built afterward can be trusted.
Show what step 6's exploration actually found: the chart types and perceptual rules that turn a dataframe of numbers into something a colleague can see at a glance.
You can clean a genuinely dirty dataset with pandas and NumPy, then show what you found in a chart that actually lands. Next up: figuring out what the patterns really mean.
Finish this section to unlock.
+100 XP
Ground the p-values and confidence intervals you'll compute next in the math that actually produces them: distributions, expectation, and how uncertainty behaves.
This is where statistics stops describing data and starts making defensible claims about it: expect to come back to hypothesis tests more than once before picking the right one feels automatic.
Apply the hypothesis testing from step 9 to a live decision: randomize an experiment, size the sample, and read the result without fooling yourself about what it proves.
You can run a hypothesis test and design an A/B test that survives scrutiny, and a few minutes of due flashcards keeps step 1's Python sharp while you build on it. Next up: turning these patterns into predictions.
Finish this section to unlock.
+100 XP
Move from describing what happened to predicting what happens next: the supervised and unsupervised algorithms, and the bias-variance tradeoff that explains why more complexity isn't automatically better.
Take this if you're heading toward economics-flavored or policy analytics work: regression diagnostics and causal identification methods that go past step 11's purely predictive focus.
Pick this up if presentation-ready statistical charts matter for your role: it builds directly on the plotting concepts from step 7 with noticeably less code.
Translate the model results from step 11 and the experiment findings from step 10 into a narrative a non-technical stakeholder will actually act on, not just nod at.
Pull every earlier step into one lifecycle view: the same cleaning, testing, modeling, and evaluation habits practiced individually, now run as the repeatable workflow a working data scientist owns end to end.
You can take a messy dataset from raw rows to a defensible conclusion: clean it, test it, model it, and explain it to someone who has never opened a notebook. That's the whole loop the Data Scientist badge certifies, not just the model in the middle.
Finish this section to unlock.
+100 XP
Finish every required step, at least 70% of them genuinely done (not skipped), to earn this badge and 500 XP.