Training Data for Machine Learning

Chapter 9. Case Studies and Stories

Introduction

Now I will share some exciting case studies and stories. I have aimed to provide a breadth of views, from engineering to annotation quality assurance. Stories from big companies to multiple sizes of startups, and education-centric data science competitions.

You may be just starting your AI path, or you may already be knowledgeable. No matter where you or your team is on your AI journey, these studies have been carefully selected to provide value and key insights. Each study will help you understand how others have implemented AI in the real world. Each study will be different, with some being more verbose and others being more brief anecdotes.

Some of the studies will be simple stories. Others will be more detailed and open, with the high-level concepts to set the level of technical depth and perspective of the scenario from the outset. In these more detailed ones, I’ll point out some of the nuances that make each example relevant and will also cover lessons learned that you can take away and incorporate into your organization.

One of the biggest recurring themes is how new modern training data is. Imagine we are at the dawn of compilers becoming common. Compilers convert high-level code to machine code, so if someone presents a way to manually optimize code,¹ it wouldn’t be very useful to a new organization that has access to compilers.

This newness brings many challenges. There are often gaps in knowledge. Case studies can be ...