Skip to main content
Engineering

Improving Uber Eats Home Feed Recommendations via Debiased Relevance Predictions

14 December 2023 / Global
Featured image for Improving Uber Eats Home Feed Recommendations via Debiased Relevance Predictions
Image
Figure 1: Smoothened empirical CVR vs. vertical position on randomized traffic
Image
Figure 2: The Examination Model
Image
Figure 3: Position Bias as a function of Vertical Position, Device OS and Feed Item Type 
Image
Image
Figure 4: Deep learning CVR model with position bias side tower. The illustration depicts the overall structure, but is not representative of the specific architecture used in production.
Charles Luo

Charles Luo

Charles Luo was an Applied Scientist Ph.D. intern at the Uber Eats search and discovery team. He is currently pursuing his Ph.D. in Biostatistics at the University of California, San Diego.

Utkan Candogan

Utkan Candogan

Utkan Candogan is a Senior Applied Scientist working on ranking and recommendations problems that surface in the Uber Eats app. Previously, Utkan received his Ph.D. degree from California Institute of Technology, specializing in mathematical optimization.

Brett Vintch

Brett Vintch

Brett Vintch was a science leader at Uber Eats, working on consumer problems like discovery, personalization, and cart building, and merchant problems like merchant selection and growth.

Posted by Charles Luo, Utkan Candogan, Brett Vintch

Category: