Machine Learning Foundations
Models that generalise, evaluated in a way that would survive an audit.
Tobias Lund
Data scientist · ex-forecasting lead
- Analytics
- Machine Learning
- Python
What you will be able to do
- Frame a business problem as a learnable task
- Build validation that reflects how the model will be used
- Detect and eliminate the leakage that inflates every score
- Pick metrics that match the decision being made
- Explain a model's behaviour to a non-technical stakeholder
About this course
The gap between a model that scores well and a model that works is almost entirely about evaluation. This course covers the core algorithms, but spends most of its time on the things that decide whether they help: leakage, validation strategy, calibration, and knowing what a metric is quietly optimising for.
Curriculum
3 sections · 9 lessons
- 1From business question to learning taskTarget definition, and the choices hidden inside it.Preview20 min
- 2Validation that mirrors realityTime-based, grouped and stratified splits.24 min
- 3Leakage: the silent score inflatorWhere it hides and the checks that find it.22 min
Before you start
- Comfortable with pandas
- High-school level algebra
Student reviews
5.0 from 4
Ryan Whitfield
19 Aug 2026Changed how my team works
I ran the middle section as a lunch-and-learn series with my team. We have since adopted three of the practices as defaults. Rare for a course to have that effect.
26 found this helpful
Hana Kimura
19 Aug 2026The instructor clearly does this for a living
Every example smells like it came from a real codebase rather than a tutorial. The asides about what goes wrong in production were the most valuable part for me.
16 found this helpful
Amara Diallo
19 Aug 2026The quiz actually taught me something
I got two questions wrong and the explanations made the reason obvious immediately. That is a rarer thing than it should be.
23 found this helpful
Daniel Osei
19 Aug 2026The instructor clearly does this for a living
Every example smells like it came from a real codebase rather than a tutorial. The asides about what goes wrong in production were the most valuable part for me.
17 found this helpful