What the work is actually typed into. Titles are on the spines; the notes underneath are the three tellings, in short — for a ten-year-old, for the architect, and for the boardroom.
6 concepts on this shelf · Beginner → Intermediate · about 2 min read
Kitchen table — The language people write instructions in. It reads almost like English, which is why so many people use it.
Whiteboard — Not fast, but it is the glue: the entire scientific and ML ecosystem assumes it, and the C and CUDA underneath do the heavy lifting anyway.
Boardroom — The default skill on any AI team. If you are hiring for AI, you are hiring Python — and the libraries matter more than the language.
Python, programming language, data science, scripting
Kitchen table — A very fast calculator for whole tables at once, instead of one number at a time.
Whiteboard — Vectorised arrays and dataframes. Learn where the copies happen, learn groupby properly, and stop writing loops over rows.
Boardroom — The spreadsheet layer of AI work — where data gets cleaned, joined and reshaped before anything intelligent happens to it.
NumPy, pandas, dataframe, array, vectorisation, data wrangling
Kitchen table — A box of ready-made learning machines. You pick one, feed it examples, and it learns.
Whiteboard — Fit, predict, transform — one consistent interface over most classical ML, plus pipelines that keep preprocessing honest between train and serve.
Boardroom — For a great many business problems this, not a large language model, is still the right and far cheaper answer.
scikit-learn, sklearn, classical machine learning, pipelines, regression, classification
Kitchen table — The workshops where the big brains get built, piece by piece.
Whiteboard — Autograd, tensors, GPUs and the training loop. Everything above them — the fine-tuning libraries, the serving stacks — is a convenience over these two.
Boardroom — Where deep learning models are actually built and trained. Relevant mostly because they decide your hardware bill.
PyTorch, TensorFlow, deep learning framework, GPU, tensors, autograd, training loop
Kitchen table — A lab diary. What you tried, what happened, so you can find your way back to the good one.
Whiteboard — Jupyter for exploration, MLflow or Weights & Biases for runs, parameters, metrics and artefacts. Notebooks are for thinking, not for production.
Boardroom — Reproducibility. If nobody can say which data and which settings produced last quarter's model, you do not have an asset, you have an anecdote.
Jupyter notebook, MLflow, Weights and Biases, experiment tracking, reproducibility
Kitchen table — The pipes that carry information from where it is kept to where it is needed, and clean it on the way.
Whiteboard — Batch and streaming, orchestration, schema drift, backfills, and the unglamorous truth that pipeline failures cause more incidents than models do.
Boardroom — The plumbing. Underfund it and every AI initiative downstream inherits the mess.
data pipeline, ETL, ELT, orchestration, streaming, data engineering, schema drift