MLOps in Real-Time Bidding

Abstract: 

This talk dives deep into the complexities of the Real-Time Bidding (RTB) ecosystem and unveils our successful approach to building a machine learning stack optimized for low-latency online predictions. We'll explore the design and implementation of each crucial component, including:

Model Server: We'll discuss how we chose and configured the model server to ensure lightning-fast delivery of model predictions within the tight timeframes of RTB auctions.

Feature Store: Delving into the feature store, we'll explain how it efficiently manages and serves the real-time features that fuel our machine learning models.

Workflow Orchestration: We'll shed light on the workflow orchestration system, the invisible conductor that coordinates the seamless flow of data and tasks throughout the machine learning pipeline.

Model Registry: The model registry plays a critical role, and we'll explore how it tracks, catalogues, and versions our models, ensuring clarity and control.

Online Model Variant Validation: Maintaining model performance is vital. We'll share our strategies for rigorous online model validation, guaranteeing that deployed models continue to deliver optimal results.

To ground these concepts in reality, we'll use the Bid Shading problem as a compelling case study. We'll unpack the challenges of bid shading in RTB auctions and outline the machine learning solution we implemented. Finally, we'll showcase the real-world impact – the millions in revenue generated since deploying this solution – to demonstrate the effectiveness of our machine learning stack in action.

Bio: 

Dr. Shuai Yuan currently serves as the Director of Data Science at Comcast Freewheel. He completed his PhD at UCL in 2015 and has since maintained a steadfast focus on advancing his expertise in computational advertising, machine learning, and information retrieval within the industry. Dr. Yuan has held leadership roles in prominent startups such as MediaGamma and Beeswax, as well as in enterprises like Comcast Freewheel. His contributions include the delivery of over 10 machine learning applications and the establishment of 3 highly effective machine learning engineering teams. In recent years, he's particularly interested in exploring and adopting MLOps tools and processes.

Open Data Science

 

 

 

Open Data Science
One Broadway
Cambridge, MA 02142
info@odsc.com

Privacy Settings
We use cookies to enhance your experience while using our website. If you are using our Services via a browser you can restrict, block or remove cookies through your web browser settings. We also use content and scripts from third parties that may use tracking technologies. You can selectively provide your consent below to allow such third party embeds. For complete information about the cookies we use, data we collect and how we process them, please check our Privacy Policy
Youtube
Consent to display content from - Youtube
Vimeo
Consent to display content from - Vimeo
Google Maps
Consent to display content from - Google