Leslie Brooks: QA For Data Science

How is QA for Data Science Different from QA for Traditional Applications?

Audience: QA engineers, data scientists, analysts, product managers, and developers exploring quality in data-driven projects

Quality Assurance in traditional application development focuses on predictable behavior, functional correctness, and well-understood user flows. But what happens when your “application” is a machine learning model, your output is a probability, and your requirements are full of ambiguity? This talk explores the unique challenges and opportunities of applying QA practices in Data Science projects. You’ll learn how the nature of data, experimentation, and modeling shifts the QA focus—and why that’s not a bad thing.

We’ll tackle questions like:

  • What does it mean to test a model’s quality?
  • Where does traditional automation break down—and what can replace it?
  • How can QA contribute meaningfully to a process where answers aren’t binary?

Then we’ll go a step further: What if we brought BDD to Data Science?

We’ll explore:

  • Writing meaningful requirements for model behavior
  • Using examples and domain language to build shared understanding
  • Automating requirements in a way that supports experimentation and accountability

Expect practical strategies, real-world lessons, and new ways of thinking about quality when the system learns, shifts, and sometimes surprises us.

This talk pairs well with the hands-on workshop: “An Introduction to BDD – Behavior-Driven Development as Your QA Superpower.”

Leslie BrooksLeslie Brooks

Leslie Brooks has decades of experience in QA and specifically in automated testing, but is always laser focused on value. Automated tests that cost too much to maintain don't deliver the value we should demand. He loves teaching about Specification by Example/BDD and the tremendous improvements it can bring to the entire organization. When we are able to write better requirements, write better code (and fewer bugs), find and fix bugs faster, and deliver better quality code faster, then he feels his work has been accomplished.