Summary
My conversation with Meri was both insightful and thought-provoking. With a wealth of experience and expertise, she is well placed to provide advice on how to develop a successful, sustainable, and scalable data science capability.
When discussing the oft-quoted high failure rates in AI projects, Meri was aligned with the common consensus throughout this book that data science and AI projects are scientific in nature, and as a result of their experimental nature, failure is part of the process. It’s important to understand that if we define failure as the project not delivering the expected business benefits, then there are a multitude of non-technical reasons for “failure” in this context. To help mitigate failures due to scalability issues, her recommendation is to start with the problem – and not the data – to build solutions that are truly scalable and aren’t limited to the dataset used for a POC or small-scale solution.
I...