Data Science

uv, a virtual environment, JupyterLab running locally, and reading the error when it breaks.

Variables, lists, functions, and loops, taken only as far as data work requires.

Why out-of-order execution quietly ruins results, and restart-and-run-all as a reflex.

Rows, columns, dtypes, and loading a CSV that fights back on encoding and headers.

Filtering, sorting, and summarizing until you can describe a dataset in five sentences.

Using an AI assistant for pandas, and the three checks that catch a column it invented.

Missing values, duplicates, wrong types, and other people's spreadsheets.

Selecting with .loc, copy-on-write in pandas 3.0, and why chained assignment no longer works.

groupby, aggregation, long against wide, and split-apply-combine as a single idea.

Merge keys, join types, and checking row counts before you believe the output.

Querying CSV and Parquet in place, and when SQL or Polars beats a chain of pandas calls.

Encodings, honest axes, and one figure that makes its point before the caption does.

Random samples, sampling variability, and simulating ten thousand coin flips to see it.

Stating a hypothesis as a process you can run, then choosing a test statistic for it.

Shuffling group labels to compare two groups without a formula or a lookup table.

Resampling one sample to get an interval, and what 95% emphatically does not mean.

The central limit theorem as something you observe in simulation, not a theorem you memorize.

Randomized against observational data, and the traps that end analyst careers.

Fitting a line, reading residuals, and refusing to call the result a prediction.

Defining a metric people cannot quietly game, and naming the ones already gamed.

Turning a notebook into a document with Quarto so a stranger can rerun it next year.

Structuring the argument, presenting it in ten minutes, and surviving the questions.