AI Model Platform Notebooks
Data scientists use “notebooks” on Capital One’s Model platform to experiment, create code, and deploy models
The business wanted associates to pause their notebooks to reduce costs
Data scientists want to have the latest data available, and pausing can disrupt active data processing
Leaving a notebook running (“unpaused”) is very expensive
I conducted a workshop with twelve users to understand what could get them to pause their notebooks
My team designed and presented two prototypes for feedback and comparison
Option 1: A pre-selected checkbox
Option 2: Radio buttons
The user feedback suggested users needed to see and make a choice, and understand the impact of that choice
Radio buttons forced users to stop and think about their choice, and to read the options carefully
The winner: Radio buttons
Computing costs decreased 27% after the change went into effect
Lessons learned:
AI modeling is a high-focus environment, so a good user experinece involves automating as much of the process as possible for the data scientists