Data Science & Analytics
Python, pandas, statistics, experiments, SQL and honest communication, practised on real and clearly labelled synthetic data.
14 chapters · 39 lessons
Data Science & AI
From your first dataset to machine learning, neural networks, generative AI and production systems: six formations you can enter where your skills already are.
Beta Every formation in this division is a beta version. Lessons, exercises and translations are still being reviewed and may change.
Six formations that share one skill graph. Data scientists ask whether the evidence supports a conclusion. Machine-learning engineers make a model generalise. Deep-learning engineers understand training from the gradients up. AI engineers make models fast, reliable and affordable in service. LLM application builders make probabilistic components behave inside real software. Data and MLOps engineers keep all of that correct every day.
They overlap, but they are different jobs. Enter where your skills already are.
Applied Machine Learning
Regression, classification, ensembles, evaluation and pipelines, implemented by hand first, then with scikit-learn.
13 chapters · 30 lessons
Deep Learning & Neural Networks
Build autodiff, optimisers, convolution and attention yourself, then train real models in PyTorch on your own computer.
10 chapters · 22 lessons
AI Engineering & Accelerated Computing
Machine learning from the mathematics up, in NumPy you can run here, then PyTorch, GPUs, inference and serving on your own machine.
12 chapters · 29 lessons
Generative AI & Large Language Models
Tokens to transformers, RAG, tools, evaluation and security, built and tested against a deterministic stand-in model in your browser.
9 chapters · 21 lessons
Data Engineering & MLOps
Modelling, idempotent pipelines, validation, orchestration, lineage, versioning, monitoring and rollback, built by hand so that tools make sense.
8 chapters · 19 lessons
What each formation is for
Data Science & Analytics
Investigate unfamiliar data and reach conclusions you can defend, including "the data can't answer this".
Builds: A reproducible investigation of a real question, with its limits stated.
Where to start: After basic Python. Start at the beginning, or at statistics if you already use pandas.
Applied Machine Learning
Frame a prediction problem, beat a baseline honestly, prevent leakage, and explain what the model can't do.
Builds: A predictive model with justified methodology and reproducible evaluation.
Where to start: After NumPy and basic statistics. The mathematics is taught where it is used.
Comes after: Data Science & Analytics
Deep Learning & Neural Networks
Understand neural networks as mathematics and computation, build their parts, and diagnose failed training.
Builds: A trained, evaluated and documented network, after writing backpropagation and attention yourself.
Where to start: After gradient descent and model evaluation.
Comes after: Applied Machine Learning
AI Engineering & Accelerated Computing
Make models fast, small, measured and dependable in service, on CPUs and GPUs, including CUDA fundamentals.
Builds: An inference service with measured performance, tests and deployment documentation.
Where to start: After Python. The early chapters are a fast route if you skipped Deep Learning; skip them if you didn't.
Comes after: Deep Learning & Neural Networks
Generative AI & Large Language Models
Build LLM applications with retrieval, tools, evaluation and security, and measure where they fail.
Builds: An original application with a golden dataset, security controls and measured limitations.
Where to start: After professional Python. The architecture chapters assume only that; the foundations chapter teaches attention at the level you need.
Comes after: Python
Data Engineering & MLOps
Build pipelines and ML operations that stay correct: validation, idempotence, lineage, versioning, monitoring, rollback.
Builds: A reproducible data-to-model pipeline with validation, tests and monitoring.
Where to start: After Python and SQL. The MLOps chapters also expect a first trained model.
Comes after: Python
Ways through the academy
Nothing locks you into an order: every chapter stays open, so skip what you already know.
- Beginner to data scientist You know basic Python.
- Machine-learning engineer You already analyse data and can code.
- Accelerated AI engineer You write solid Python, and C++ helps for the CUDA chapters.
- LLM application builder You are a developer who wants to build with language models, properly.
- Data engineer You like reliable systems more than charts.
What completion means, honestly
Finishing a formation means you completed its activities. It is not a certificate, and Understand issues none for this division today. Projects are assessed by you against a published rubric until human review exists.
- Lesson viewed Tracked
You opened the lesson. This shows exposure and nothing more. - Exercise attempted Tracked
You ran the checks at least once. - Exercise completed Tracked
The checks passed in your browser. Browser checks can be bypassed. - Skill demonstrated Tracked
Activities on the skill right first time, without help, and again at least seven days later. - Project completed Self-reported
You ticked the rubric yourself. Nobody checked it. - Project reviewed Not available yet
A person applies the rubric to your repository. You can submit a link - Competence independently verified Not available yet
A supervised assessment with identity checks. Not offered yet.