Python for data science — pandas, polars, real datasets
Data science books love arrays like [1, 2, 3]. Hiring managers care if you can groupby a real dataset and explain what you did. We optimise for the second.
The standard 'Python for data science' funnel asks you to install Anaconda (8 GB), figure out conda vs. pip, fight matplotlib backends on macOS, and finally — three hours in — write your first `pd.read_csv`. Two-thirds of the people who started don't make it to that cell. Of those who do, half learn pandas patterns that look right in a tutorial but fall apart on a 2 GB file in a real job.
Our Data Science Applied track is 100 lessons of writing code against real CSV and JSON files in the browser, with pandas first (because every interview asks about it) and polars second (because fast modern teams ship in it). DuckDB for SQL, matplotlib + plotly for charts, one end-to-end mini-project that joins three tables and reports a business metric. The AI tutor catches the bug before you re-run the notebook cell — guiding question, not the answer.