This episode features Shreya Shankar, a PhD student at UC Berkeley and former ML engineer at Viaduct, Google Brain, and Facebook. Shreya joins Hugo Bowne-Anderson to discuss key insights from her team’s paper Operationalizing Machine Learning: An Interview Study, which uncovers common practices and pain points experienced by ML practitioners. They explore the main tasks in the ML production lifecycle, strategies for sustaining model performance, and future opportunities in ML Engineering tools and research.

Shreya and Hugo cover:

• Main tasks involved in the ML production lifecycle
• Factors that determine the success of ML deployments
• Pain points of deploying ML models to production and how to address them
• Strategies for maintaining model performance in production

_{This episode was recorded as a fireside chat on 10/27/2022.}