Beyond MLOps: A Closer Look at Operational Machine Learning & Why it Matters
While critical aspects of MLOps like infrastructure, CI/CD pipelines, logging and monitoring are much discussed in the machine learning world today, there has been less of a focus on the actual operations of machine learning. This talk will explore key topics and questions related to the operational considerations of developing and deploying machine learning models in real-world environments, including: how to respond when something goes awry; data-centric approaches to model development and how to course-correct for suboptimal training data; and what happens when novel external environments (e.g. a global pandemic) create outputs that run counter to developers’ expectations.
Zachary Hanif is Capital One’s Vice President of Machine Learning, currently working to democratize and de-risk access to modeling capabilities across a data-driven enterprise. He has a career background focused on applications of machine learning and advanced modeling innovation, research, and product delivery across diverse problem domains in startups and large enterprises. Zachary is passionate about identifying novel uses of machine learning and large-scale distributed systems technologies that generate measurable value, pragmatic innovation, and ensuring that technical solutions, particularly those leveraging AI/ML, are responsible, ethically sound, and robust to challenge.