Driving ML Data Quality with Data Contracts

Explore the concept of Data Contracts and their role in improving data quality and reliability for machine learning models in this 35-minute talk by Andrew Jones, Tech Lead at GoCardless. Learn how empowering data consumers to collaborate with data generators can lead to more effective ML models and data-driven products. Discover the key ML models used at GoCardless, the challenges of poor quality data, and the implementation of data contracts as a solution. Gain insights into creating good quality data, treating data as an API, and changing data culture within an organization. Understand the importance of aligning on problems, finding tailored solutions, and recognizing the value of data in driving successful ML initiatives.

Syllabus

[] Musical introduction to Andrew Jones
[] Andrew's background
[] Driving ML Data Quality with Data Contracts
[] GoCardless
[] The Key ML Models at GoCardless
[] Data is critical to a model's performance
[] The data platform at GoCardless in 2021
[] Ultimately, we believe that data is of poor quality
[] There must be a better way...
[] What is good quality data
[] An API for data?
[] Introducing data contracts
[] What is data contract?
[] An example data contract
[] Isolated GCP projects
[] It's not really about the implementation...
[] Align on the problem
[] Work out how best to solve it for us
[] Change the data culture
[] Our data has value