Data Governance for Data Science with Adam Wood - #578

Feature Stores

Feature stores play a crucial role in data science by maintaining consistency and enabling feature reuse. Adam Wood emphasizes the importance of making curated feature sets readily available for data scientists to use, ensuring agility and efficiency in developing new use cases 1. He compares the process to cooking a meal without a recipe, highlighting the need for detailed metadata management to avoid starting from scratch each time 1.

We want other people to start consuming that right away.

--- Adam Wood

Sam Charrington adds that feature stores act as a marketplace, promoting reuse and preventing teams from reinventing the wheel 2.

   

Feature Reuse

Promoting feature reuse across teams enhances efficiency and reduces redundancy. Adam Wood discusses how Mastercard collaborates with data science teams to certify top-quality assets, making them easily accessible for reuse 2. This approach not only saves time but also ensures consistency in data usage across projects.

We can't guarantee reuse, but we can strongly promote reusable feature sets.

--- Adam Wood

Sam Charrington highlights the importance of profiling data to understand its biases and ensure responsible use in machine learning applications 3.