
Make an analysis anyone can rerun: a notebook that runs top to bottom in a fresh kernel, reads its data by a relative path rather than one analyst machine, fixes its seed so a sampled spot check does not move between runs, and writes its answers to a file instead of leaving them in cell output.
Load an extract that arrived in three formats without letting the reader guess: declare each column type from the shipped schema, keep zero-padded codes as text, skip the preamble and trailer of a fixed-width file, and record a profile per file with shapes, types, missing and distinct counts and the summary statistics.
Give every new extract the same first look before anyone charts it: a fresh-kernel run that loads three formats with explicit types and writes a profile per file, raising the defined red flags - columns past the missing threshold, numbers still read as text, and columns that never vary - while leaving clean columns unflagged.
Analyse data hands-on in Jupyter notebooks with Python or R: reproducible notebooks, typed loading, profiling, cleaning, aggregation and reporting.
1 module: Notebooks & Data Loading.
3 exercises, all easy.
Challenges run on Jupyter Notebook.
No. Every challenge runs in a browser-based IDE with everything already set up, and your code is graded automatically against a test suite.