Sourabh Bansod
Sourabh Bansod is a Senior Engineering Manager at Google, where he leads the technical strategy and execution for YouTube Shorts recommendations. His team builds large-scale ML systems for retrieval, ranking, and content discovery, powering creator inspiration, cultural trends, and billions of daily recommendations. Previously, Sourabh led ML models for conversion and retention that scaled YouTube Premium and Music from launch to 50M subscribers. He holds multiple patents, has published at ACM RecSys, and contributes as a program committee member for top AI and recommender systems conferences.
Research Areas
Authored Publications
Sort By
Streaming Trends: A Low-Latency Platform for Dynamic Video Grouping and Trending Corpora Building
Scott Wang
Caroline Zhou
Ashkan Norouzi Fard
Yongzhe Wang
Yaping Zhang
Li Zhang
Mingyan Gao
Qiao Zhang
2025
Preview abstract
This paper presents Streaming Trends, a real-time system deployed on a short-form videos platform that enables dynamic content grouping, tracking videos from upload to their identification as part of a trend. Addressing the latency inherent in traditional batch processing for short-form video, Streaming Trends utilizes online clustering and flexible similarity measures to associate new uploads with relevant groups in near real-time. The system combines online processing for immediate updates triggered by uploads and seed queries with offline processes for similarity modeling and cluster quality maintenance. By facilitating the rapid identification and association of trending videos, Streaming Trends significantly enhances content discovery and user engagement on the YouTube Shorts platform.
View details
Preview abstract
In this paper, we document the development of a recommender system that provides inspiration to existing content uploaders and new future content uploaders to encourage participation. Our contributions are two-fold: 1) Inspiration Framework: We present a novel framework for building a recommender system that goes beyond traditional consumption-focused metrics, specifically addressing the need for creative inspiration to lower barriers for participation. This framework is adaptable in the design of large-scale recommender systems in other domains. 2) Empirical Evaluation: We conduct systematic evaluation via live experiments to prove the values of the proposed system in increasing daily participation and participants.
View details
Preview abstract
Multi-task prediction models and value models are the de-facto standard ranking components in modern large-scale content recommendation systems. However, they are typically optimized to model users’ passive consumption behaviors, and rank content in a way to grow consumption-centric values. In this talk, we discuss the key insight that it is possible to grow content generation (from users) through a new ranking system and its ecosystem value. We made the following key technical contributions in this system: (1) introducing ranking for content generation based on a categorization of user participation actions of different sparsity, including proxy intent action or access point clicks. (2) improving sparse task prediction quality and stability by causal task relationship modeling, conditional loss modeling and ResNet based shared bottom network. (3) personalizing the value model to minimize conflicts between different values, through e.g. ranking inspiring content higher for users who actively generate content. (4) conducting systematic evaluation of proposed approach in a large short-form video UGC (User-Generated Content) platform.
View details
