Certification programme
The algorithms, properly understood. You implement the core models before reaching for a library, so you know why one fails and another works.
What you will be able to do
Full syllabus
9 modules · 47 topics · 184 concepts
Types of learning
The core vocabulary
The fundamental trade-off
Evaluation discipline
When not to use ML
You can state any problem as a learning task and identify what could go wrong.
Linear algebra
Calculus for optimisation
Gradient descent
Probability
Statistics for ML
You can read an algorithm description and follow what it is optimising.
Exploratory analysis
Missing data
Encoding
Scaling
Feature engineering
Pipelines
You can build a preprocessing pipeline that does not leak information.
Linear regression
Assumptions
Polynomial and non-linear
Regularisation
Regression metrics
You can fit, regularise and interpret a regression model.
Logistic regression
K-nearest neighbours
Naive Bayes
Support vector machines
Decision trees
Classification metrics
You can build a classifier and choose a threshold defensibly.
Why ensembles work
Bagging and random forests
Boosting
Stacking
Interpretability
You can train and tune a gradient-boosted model and explain its predictions.
Clustering
Evaluating clusters
Dimensionality reduction
Anomaly detection
Association rules
You can cluster and reduce dimensionality, and judge whether the result is meaningful.
Cross-validation
Hyperparameter search
Imbalanced data
Model selection
Common failures
You can tune a model honestly and report a number you would defend.
Persisting a model
Serving
Monitoring
Experiment tracking
Ethics and risk
You can serve a model behind an API and monitor it for drift.
5 builds you keep
A churn problem on imbalanced data where the business cost of a false negative is far higher than a false positive. The work is in threshold selection and explaining the trade-off, not in fitting the model.
Build a model that decides, then use SHAP to explain any individual rejection in plain language. Includes testing whether the model behaves differently across groups it should not.
The project that teaches why random cross-validation lies. Proper time series splits, a seasonal baseline you have to beat, and an honest error figure.
Almost no labels, heavy imbalance, and a cost to investigating a false alarm. An unsupervised approach measured against the handful of confirmed cases.
Trained pipeline, FastAPI endpoint, input validation, and a monitor that flags when incoming data stops looking like the training set. The part of ML that interviews ask about and courses skip.
Hands on with
Before you enrol
Less than people fear, more than zero. You need to be comfortable with algebra and willing to follow a derivative. The second module builds exactly what is needed. If you can follow a gradient descent update, you can follow this course.
This first. Most real business problems are tabular, and gradient boosting beats a neural network on tabular data more often than not. Deep learning matters for images, audio and text, which is our AI course.
Because when a model behaves oddly, the people who can fix it are the ones who know what it is doing. You implement the core few, then use the library for everything else, which is also how the job works.
No. Everything in this course runs on a normal laptop. GPUs matter for deep learning, and where we need one we use a free cloud notebook.
On its own, no, and we would rather say so. What gets interviews is a project you can explain end to end, including what you tried that did not work. That is why the projects here are built around judgement calls rather than around accuracy scores.
Where this leads
One skill on a CV is a hobby. These are the courses students take alongside this one, in the order that builds an actual role.
From neural network fundamentals to working with large language models. An applied AI programme for people who want to build systems, not just talk about them.
The complete analyst toolkit in one programme: spreadsheets, SQL, visualisation and Python, applied to real datasets and finished with a dashboard you can defend in an interview.
What happens beneath your program, taught at the Linux terminal so the theory stays concrete. Doubles as the foundation for cloud and cybersecurity work.