Machine Learning Roadmap

A step-by-step Machine Learning roadmap: what to learn in order, the related skills and the best places to learn each step.

The stack, end to end

  • Data: pandas, NumPy, Data cleaning, Jupyter notebooks, Data visualisation, Descriptive statistics, Probability, Hypothesis testing, Correlation
  • AI and ML: Machine learning, scikit-learn, Model evaluation, Feature engineering, Neural networks, PyTorch, TensorFlow, Computer vision, Natural language processing

  1. Python for data: pandas and NumPy

    Suggested pace: Start here

    Load, clean, reshape and summarise data with pandas and NumPy, and draw charts with Matplotlib or similar. Notebooks let you show your working.

    Skills

    • pandas
    • NumPy
    • Data cleaning
    • Jupyter notebooks
    • Data visualisation

    Tips

    • Pick a public dataset you care about and publish a short write-up of what you found.

    Where to learn

  2. Statistics for data work

    Suggested pace: Next

    Learn averages and spread, distributions, sampling, correlation, hypothesis tests and how not to be fooled by a pattern that is only chance.

    Skills

    • Descriptive statistics
    • Probability
    • Hypothesis testing
    • Correlation

    Tips

    • Always ask: how many data points is this based on?

    Where to learn

  3. Machine learning fundamentals

    Suggested pace: Then

    Learn how models learn from data: supervised and unsupervised learning, training and test sets, overfitting, metrics, and how to use scikit-learn.

    Skills

    • Machine learning
    • scikit-learn
    • Model evaluation
    • Feature engineering

    Tips

    • Spend more time understanding the data than choosing the model.

    Where to learn

  4. Deep learning

    Suggested pace: After that

    Neural networks power image, speech and language models. Learn the ideas, then use PyTorch or TensorFlow on a small project.

    Skills

    • Neural networks
    • PyTorch
    • TensorFlow
    • Computer vision
    • Natural language processing

    Tips

    • Re-use pre-trained models before training your own.

    Where to learn

Frequently asked questions

Where should I start with Machine Learning?

Start at step 1 and work down in order, because later steps build on earlier ones. If you are signed in, the roadmap compares your resume with each step and marks where you could start.

Is Machine Learning worth learning?

Check how many job adverts in your area mention Machine Learning, and whether it fits the kind of work you want. This roadmap shows what to learn; the Browse jobs page shows what employers are asking for.

How do I know I am ready to apply?

When you can build a small project from start to finish without copying a tutorial, and explain why you made your choices. Apply while you keep learning; you do not need to know everything.