Machine Learning
From clean datasets to deployed models — with the statistical honesty the field demands.
- Mode
- Online
- Level
- Intermediate
- Certificate
- Verifiable
- Stack
- 6+ tools
Machine learning roles reward people who can be trusted with a number — and punish people who can't explain why their model works. This program builds that trustworthiness, from validation discipline to production monitoring, so your results hold up under scrutiny.
What this program actually does.
It's easy to get a model to score well on a dataset and much harder to know whether that score means anything. This program is built around that honesty: validating properly, understanding what a metric actually tells you, and being suspicious of results that look too good.
You'll take models past the notebook — packaging, serving and monitoring them the way they behave once real, changing data hits them in production.
Who should join
- Students with basic Python
- Analysts moving into modelling
- Engineers adding ML to their toolkit
- Data professionals preparing for ML interviews
Roles this prepares you for
Structured phases, built to compound. Here's exactly what you'll learn.
Data work
Cleaning, features, leakage and validation strategy.
- Engineer features that actually help a model, not just add noise
- Catch data leakage before it quietly inflates your accuracy
- Design a validation strategy that mirrors how the model will be used
Core models
Regression, trees, ensembles and model selection.
- Know when a simple regression beats a complex model, and why
- Use tree-based models and ensembles correctly, not by default
- Select a model based on the problem, not on what's trendy
Evaluation
Metrics that match the business problem, error analysis.
- Choose the metric that actually matches the business cost of errors
- Do error analysis that tells you what the model is getting wrong, and why
- Explain a model's limitations honestly instead of overselling it
Unsupervised methods
Clustering, dimensionality reduction and anomaly detection.
- Cluster data to find structure nobody labelled for you
- Use dimensionality reduction to explore and visualise high-dimensional data
- Build a simple anomaly detector for a real-world use case
Delivery
Packaging, serving and monitoring model drift.
- Package a model so it's easy to version and deploy
- Serve predictions through a real API with sane latency
- Monitor for data and model drift instead of finding out from users
MLOps basics
Experiment tracking, reproducibility and retraining pipelines.
- Track experiments so results are reproducible, not lucky
- Version datasets and models the way you'd version code
- Set up a retraining pipeline that keeps a model current over time
Technology stack
Projects
- Forecasting model with validated baseline
- Classification service with monitoring
- Customer segmentation using unsupervised methods
- End-to-end ML pipeline with retraining
Learning outcomes
- Build models that generalise
- Explain results to non-technical stakeholders
- Choose the right evaluation metric for the problem
- Ship and monitor a model in production
Online · Live mentorship · Intermediate
Machine Learning questions.
Ready to start the Machine Learning track?
Talk to our team about the track, internship or institutional program that fits where you are today.