Speakers

Saurav Raj (ML Engineer, PayPal)
Sharmili Srinivasan (ML Engineer, PayPal)

Abstract

Machine learning inference pipelines are getting complex as they often comprise multiple models to make a prediction. These models typically follow patterns such as chaining, fanout, or ensemble in production. A streamlined model deployment mechanism that supports these patterns would reduce the effort of ML Engineers and make the ML systems more agile. In this talk, we will learn about inference graphs in ML systems and use them to design and deploy ML pipelines that require composing multiple ML models.

Slides

https://docs.google.com/presentation/d/1dTTDEbgE9oMd5h_oB3vfYbM-sZ1eRsPUC6MhPVU8RxA/edit?usp=sharing