Back

Speaker "Sam Mefford" Details Back

 

Topic

Engineering for Machine Learning Predictions

Abstract

In this session you can learn about what it took to run deep-learning models at production scale to extract Obituaries from 500 Million Newspapers pages. We will discuss what makes getting model predictions in production challenging, and why the state-of-the-art is very immature. Without technical details, we will review the AWS components used, challenges faced, and lessons learned. We will review cost saving practices and discuss the collaboration required between Data Scientists and Engineers.
Who is this presentation for?
Engineers, Data Scientists, and Business People who want a deeper understanding of what it takes to run batches of dta through Machine Learning models to get predictions in production.
Prerequisite knowledge:
General technical knowledge. Some familiarity with what a Machine Learning model is.
What you'll learn?
In this session you can learn about what it took to run deep-learning models at production scale to extract Obituaries from 500 Million Newspapers pages. We will discuss what makes getting model predictions in production challenging, and why the state-of-the-art is very immature. Without technical details, we will review the AWS components used, challenges faced, and lessons learned. We will review cost saving practices and discuss the collaboration required between Data Scientists and Engineers.

Profile

Sam Mefford is a senior member of the Machine Learning Engineering and Architecture team at Ancestry.com. Sam has over 20 years experience working as an engineer. He is passionate about using cloud technology (primarily AWS) to run machine learning prediction pipelines at scale in a cost-effective way. Sam has worked with many technology stacks and tools, such as Java, RDBMS, NoSQL, and Search, among others. He also likes sharing his knowledge with other engineers and business leaders. His presentations have been noted for their clarity and depth.