Data Science & Machine Learning

What data scientists do: the workflow

See the six stages of a data science project, from framing a business question to deploying a result, using one worked example from a coffee shop. Learn where time really goes and the mistakes that sink beginners.

Free notebooks: Google Colab and Kaggle

Set up a free data science workspace in your browser with Google Colab and Kaggle Notebooks. Run code, bring in your data, install packages, use a GPU only when needed, and avoid the habits that lose work.

Python for data: NumPy and pandas

Learn the two libraries behind Python data work. Load a CSV, find and fix missing values, filter and sort rows, join tables and summarise sales with groupby, using a small coffee shop orders file.

Exploratory data analysis and charts

Explore a year of coffee shop sales end to end: check data quality, spot outliers, make histograms, boxplots, scatter plots, a correlation heatmap and time charts, then write down what you found before modelling.

Statistics you actually need

Learn the statistics a data scientist uses every week: mean vs median, spread, the normal curve, confidence intervals, p-values and A/B tests, plus the traps (correlation, small samples, p-hacking) that make results wrong.

How machine learning learns: supervised vs unsupervised

Understand what machine learning really is: learning rules from examples by reducing error. Learn features and labels, regression versus classification, supervised versus unsupervised learning, and why overfitting is the main risk.

Regression: predicting numbers

Build your first regression model with scikit-learn: split data, beat a baseline, train linear regression, read its weights, measure error in real units with MAE, RMSE and R2, check residuals and compare with a random forest.

Classification: predicting categories

Build your first classifier with scikit-learn: split the data, set a baseline, train logistic regression and KNN, read a confusion matrix and use probabilities. You will see why scaling and a baseline matter.

Train, validation, test and overfitting

Learn how to measure honestly whether a model will work on new data: train, validation and test sets, cross-validation, overfitting and underfitting, and data leakage, with runnable examples.

Metrics: accuracy, precision, recall, F1 and ROC

Accuracy can hide a useless model. Learn the confusion matrix, precision, recall, F1, thresholds and ROC AUC with runnable examples, and how to pick the right metric for fraud, spam or screening problems.

Feature engineering

Turn raw columns into features a model can use: dates, ratios, log transforms, missing values and categories. Build a ColumnTransformer pipeline and prove with cross-validation which new features help.

Decision trees, random forests and gradient boosting

Learn how decision trees, random forests and gradient boosting work, why trees overfit, and how to compare and tune them in scikit-learn. These are the standard models for tabular data.

Clustering and dimensionality reduction

Find groups in unlabelled data with k-means, choose k using inertia and silhouette, describe the clusters in plain words, and use PCA to compress many columns into a few and draw them.

Project: predict customer churn

Build an end-to-end churn model: explore the data, set a baseline, compare logistic regression and gradient boosting with cross-validation, choose a threshold by business cost, explain the drivers and save the model.

Neural networks explained

Understand how a neural network works: neurons, layers, activation functions, loss and gradient descent. You will run tiny NumPy examples so the maths feels concrete, ready for PyTorch next.

Deep learning with PyTorch: your first model

Build and train your first neural network in PyTorch on a real dataset: tensors, DataLoader, model, loss, optimiser, the training loop, evaluation, saving and using a GPU. Runs free in Colab or Kaggle.

Computer vision with CNNs and transfer learning

Learn how convolutional networks see images, then build an image classifier by reusing a pretrained ResNet with transfer learning. Free to run on a Colab or Kaggle GPU.

NLP and transformers with Hugging Face

Use Hugging Face to run sentiment analysis and zero-shot classification, build semantic search with embeddings, and fine-tune a small transformer on your own labels, all free in Colab or Kaggle.

How AI is changing data science

See which parts of data science AI assistants speed up, where they fail silently, how the job is shifting, and how to use AI to learn faster without weakening your fundamentals.

MLOps basics: from notebook to deployed model

Take a model out of the notebook: save it with its preprocessing, serve it with FastAPI, test it, package it with Docker and learn what to monitor once it is live.