Скачать или смотреть Data Pipeline Foundations for Production Machine Learning

The scrappy, modern data scientist is raised in the discipline with little more than a laptop and an internet connection. While an ad hoc workflow is great for exploratory analysis and winning Kaggle competitions, it’s a poor fit for the needs of production machine learning. The cliche that data scientists spend 80 percent of their time finding, cleaning, and reorganizing data is at least partially due to bad habits and a lack of modern data pipelines.

In this talk, we’ll discuss the modern data pipeline as a continuously improving process flow that encompasses data ingest, data transformation, model training and deployment. In this vision, data scientists work with data engineers to create and evolve data pipelines and create shared data assets. Data cleaning, delivery and transformation is collaborative and automated so data scientists can spend more time identifying new opportunities and delivering value to customers and the business.

Joe Reis and Matt Housley from Ternary Data will be speaking