Predicting fuel efficiency is harder than it looks. MPG depends on complex, non-linear interactions between engine design, vehicle size, power, and drivetrain choices. In this project, I built and compared eleven models on a dataset of 1,500 passenger cars, moving from linear regression to gradient boosting. A tuned XGBoost pipeline achieved 5.4% MAPE, reducing error by roughly one third and providing a robust MPG prediction engine for unseen vehicles.

Lately, we have been spending more time outdoors and filming it along the way. This short note is about documenting places as they feel in real life, from quiet beaches in Texas to city walks and national parks.

A new eCommerce course is now live on afateev.co. The price went up after December 1, but you can still get 50% off by writing to us on Twitter / X.