A Crisis in Improvisation

Andreas Granström

2018-11-02
Tags:data-sciencemachine-learningprocess

Build Simple Models (or Keep it Simple, Stupid)

Don’t look for tricky ways to make your life more interesting.

I’ll keep this post simple. When you think you need Artificial Intelligence or machine learning, you actually might. But there’s a chance you don’t.

Don’t waste time building things you don’t need.

Machine learning should help your business, not hamper it. Consider that no system exists by itself, or sits in a vacuum. Decision systems will require process. Tech systems will need integration. All systems need to be maintained.

Before building, validate. Is Apache Kafka and Spark clusters the best starting point to build your recommendation system? How about you manually curate static recommendations based on some general parameter, such as country or age, and serve that as your Version 0.1 Alpha?

You can always scale up. Start small, build fast, validate your ideas and think long term. You’ll get there. Have faith.

Too many AI and machine learning projects struggle because they take too long to develop, or they become too complicated to maintain (data dependencies are hard to manage), or they weren’t needed in the first place. At worst the system works but doesn’t fit the business.

Build Simple Models, and keep things simple.