Speaker: Henry Humadi -- Machine Learning Director, Intelligence Engagement at Leanplum

Title: From notebooks to data products

Bio
Henry Graduated from McMaster University/McGill University in Canada with a Ph.D. in computational condensed matter physics. Henry has published multiple articles in peer-reviewed journals & his background in simulation/analysis of physical system has given him the tools to deal with large data sets. In the last five years, he has worked in different data science roles at various Ad-tech/martech companies. Currently, at Leanplum, Henrys leading a team that focuses on building data products to enable marketers to increase their end user engagement via Machine Learning & A/B testing.

Abstract

The transition from notebooks & scripts to a full-fledged data product takes a lot of effort. The majority of the issues tend to be engineering in nature. Even with a great data science team, most of the initial gains come from the industry knowledge, simple heuristics, clean & consistent data for insightful analytics, & not from machine learning algorithms. In this talk, we will discuss how to build a data product from scratch & what are key components such as developing a robust data pipeline, data consistency/quality, starting with a simple baseline, analytics/heuristics, & considering scalability problems to name a few. The success of the data science team & driving value from data depend on building a platform that enables experimentation & reduces the friction between the raw data & the business value.