What it was
Between April and July 2024 I worked through the 10 Academy programme: one real-world project a week, each ending in a report. The set covered sales forecasting on the Rossmann store data, time-series analysis of Brent oil prices, fraud detection on e-commerce transactions, credit scoring built from e-commerce transaction data, a data warehouse for medical business data scraped from Telegram channels, insurance claims analysis with hypothesis testing and a predictive model, a marketing analytics dashboard, exploratory analysis of solar farm data in Streamlit, and a telecom customer analysis. Most weeks ended with a Medium write-up, and the warehouse week is where I first used dbt and a proper ELT layout.
What I learned
Repetition under a deadline built habits: start from the business question, profile the data before modelling, keep the pipeline reproducible and write the report as you go. The fraud and credit scoring weeks taught me how much class imbalance and leakage shape a model's apparent quality. The warehouse week mattered most in hindsight: scraping, loading, transforming with dbt and serving a dashboard was a small version of the work I now do in production.