The Knowledge Center

Machine learning explained for people who build things. Not a random collection of blog posts — a structured learning path that takes you from "what is ML?" to "I can evaluate and apply a model," one concept at a time.

Every article is written with engineers and technicians in mind. Practical, grounded, and deliberately free of hype.

A roadmap, not a blog

The Knowledge Center is organised into four stages. Each one builds on the previous, so the concepts stack up in the right order. If you're completely new to ML, start at Stage 1. If you already know the basics, jump to wherever it gets interesting.

Stage 1 gives you the mental model — the vocabulary and core ideas you need before anything else makes sense.

Stage 2 introduces the actual algorithms — the tools you'll use to build models.

Stage 3 teaches you how to honestly evaluate whether your model is any good.

Stage 4 covers the more advanced concepts you'll encounter in the real world.

The fundamentals

What is this whole ML thing? — The vocabulary and core concepts you need before anything else makes sense. Read these five articles and you'll walk away with a working mental model.

What Is Machine Learning?

A system that learns rules from data instead of you writing the rules. Like showing it 500 examples and letting it figure out the threshold itself.

Features and Labels

The inputs and outputs of an ML model. Your process parameters are features, the thing you're predicting is the label. That's the whole idea.

Training Data vs Test Data

Why you split your data in two, and why evaluating on the data you trained on is like marking your own exam.

Supervised vs Unsupervised Learning

Supervised: you have the answers and you're teaching the model. Unsupervised: you're looking for hidden patterns. One is curve fitting, the other is clustering.

Regression vs Classification

Two flavours of supervised learning. Regression predicts a number (what will the stress be?). Classification predicts a category (pass or fail?).

The algorithms

How do the basic algorithms work? — Now you know the landscape, here are the actual tools. Starting with the simplest and building up to the workhorse of tabular ML.

Linear Regression — the ML Starting Point

Fitting a line through data points. Engineers already do this in Excel — ML just formalises it and scales it up. Includes: how the model minimises prediction error (the cost function).

Overfitting and Underfitting

The most important concept in all of ML. Your model memorised the noise, or your model is too simple. Like fitting a 20th-degree polynomial through 10 data points.

What Is a Decision Tree?

An algorithm that predicts by asking yes/no questions about your data. Completely transparent — like a troubleshooting flowchart you can read end to end.

What Is a Random Forest?

Train 100 different decision trees on slightly different data subsets and let them vote. The wisdom of the crowd, applied to algorithms.

What Is Gradient Boosting (XGBoost)?

Train trees in sequence — each one fixes the mistakes of the previous. XGBoost is the workhorse of tabular ML and the algorithm behind countless Kaggle wins.

k-Nearest Neighbours

The simplest algorithm: find the most similar data points and average their values. Like asking "what happened last time we had similar conditions?"

Evaluating your model

How do I know if my model is any good? — You can train a model now. But can you tell whether it's actually useful? These articles teach you to evaluate honestly.

Cross-Validation — Honest Model Testing

A single train/test split isn't reliable enough. Cross-validation rotates through multiple splits — like repeating a physical test and reporting the average.

Accuracy Is Not Enough

Why 95% accuracy can be meaningless when 95% of your parts are good anyway. Precision, recall, and the confusion matrix — through a defect detection lens.

Feature Importance — Which Inputs Actually Matter?

How tree-based models reveal which of your 20 process parameters actually drive the prediction. Gold for engineering root-cause analysis.

Advanced concepts

What about the more advanced stuff I keep hearing about? — With the foundation in place, these are the concepts that come up in real projects, headlines, and conversations with data scientists.

What Is a Neural Network?

Layers of simple mathematical functions chained together. Powerful for images and signals, but often overkill for spreadsheets. An honest take on when you need one.

Deep Learning vs Classical ML — When to Use Which

The payoff article: if your data is a table, start with trees. If it's an image or waveform, consider deep learning. A clear decision framework.

What Is a Surrogate Model?

Run a few expensive simulations, train an ML model on the results, use the fast model for design exploration. The concept that makes ML uniquely valuable for engineers.

SHAP — Explaining Any Model's Predictions

How to answer "why did the model predict this?" for any individual prediction. Essential for engineering trust and regulatory contexts.

The ML Project Checklist for Engineers

Not an algorithm — a decision framework. Do I have enough data? Is this actually an ML problem? What metric matters? The capstone that ties everything together.

Coming soon

Prefer to learn in a classroom?

We're developing a hands-on, multi-day ML course specifically for engineers and technicians in Belgium and the Netherlands. Small groups, practical exercises with real engineering data, and zero prerequisites in data science.

Want to be the first to know when dates are announced?

Keep me posted