Practical AI
11 lessons where you build real machine-learning models in Python from scratch, running right here in your browser: sales forecasting, churn prediction, customer segments and measuring accuracy. Then see how AI is built into ERP systems and Power BI, and how to use it responsibly.
Lessons
Machine learning at work: what does it solve?
Classification, regression and clustering, and when you don't need ML at all.
Data: features, target and splitting
Features versus the target, why we split data into training and test sets, and memorising instead of learning.
Your first model: forecasting sales with a straight line
Linear regression from scratch in Python: slope and intercept, then a sales forecast.
How wrong is the model? Measuring error
Error per prediction, mean absolute error (MAE) and root mean squared error (RMSE).
Classification: who is about to leave?
k-nearest neighbours from scratch to predict customer churn, and why scaling matters.
Is the model any good? The confusion matrix
Accuracy, precision and recall, the imbalanced-data trap, and choosing a decision threshold.
Customer segments: k-means clustering
Learning without answers: k-means step by step, and turning segments into marketing decisions.
Real tools and the ML project life cycle
pandas, scikit-learn, Jupyter and Colab, no-code tools, and the CRISP-DM method.
AI inside ERP systems
Reading invoices, demand forecasting, spotting odd journal entries, and assistants and agents in SAP, Oracle and Odoo.
AI in business intelligence and Excel
Copilot in Power BI, forecasts and anomaly detection in charts, Forecast Sheet, and preparing data for AI.
Responsible AI and your next project
SDAIA's AI ethics principles, bias, transparency, model monitoring, and project ideas.