Start Here

This short course is designed to feel like a guided handbook, not a stack of lecture notes. Start here to understand the scope, the rhythm, and what to focus on.

Who this course is for

This material is a good fit if you:

  • work with structured datasets such as transactions, surveys, operations data, customer records, or measurements
  • want a practical overview of common ML methods without turning this into a purely mathematical course
  • need enough fluency to build models, review model choices, or collaborate well with technical partners

What this course covers

The course centers on a typical tabular ML workflow:

  1. define the task and choose the right problem framing
  2. inspect the data and prepare it for modeling
  3. evaluate models in a disciplined way
  4. compare classical and modern model families
  5. turn model-building into a repeatable pipeline

What this course does not try to do

  • it does not cover deep theory in probability or optimization
  • it does not aim to make neural networks the default answer for every tabular problem
  • it does not treat AutoML as a substitute for judgment

Instead, it aims to make you effective with the kinds of tabular data problems that appear constantly in applied work.

Suggested study rhythm

  • Read one chapter at a time.
  • Keep a small working notebook or scratchpad beside you.
  • After each chapter, write down one modeling decision you would now make differently.
  • Use the mini-project to consolidate the whole workflow.

Setup suggestions

If you want to work hands-on while reading, a minimal setup is enough:

  • Python
  • pandas
  • scikit-learn
  • matplotlib or seaborn
  • a notebook environment such as Jupyter

Success criteria

By the end of the course, you should be able to:

  • translate a practical question into a supervised or unsupervised ML task
  • explain why a metric, split strategy, or preprocessing step was chosen
  • compare at least three model families for the same dataset
  • build a baseline workflow that is simple, reproducible, and defensible

Before moving on

Use this quick self-check:

  • Can I explain the difference between a prediction task and a product decision?
  • Do I know what kind of tabular data problems I care about most?
  • Am I willing to prioritize clean baselines over premature complexity?

If yes, move to the next chapter: Workflow and Problem Framing.

Next