DataFunSummit 2023: Recommendation Systems Online Summit

The DataFunSummit 2023: Recommendation Systems Online Summit will be held on August 26-27, 2023. This summit covers eight forums including core architecture, model innovation, engineering architecture and training inference, best practices, large models and recommendation, graphs and recommendation, recommendation cold start, and multi-scenario multi-task.

The summit features technical experts from Xiaohongshu, including Feng Di (Vice President of Technology), Miao Tai (Search Promotion Engineering Director), Ban Chao (Recommendation Engineering Architecture Director), Jing Lun (Machine Learning Engine Architecture Director), and Li Mu (Graphic Information Flow Algorithm Director). They will share best practices in recommendation system technology.

Three key presentations from Xiaohongshu include:

1. "Xiaohongshu's Graph-Based Business Architecture Practice in Complex Search Promotion Scenarios" by Ban Chao on August 26, 10:15-10:50 in the "Core Architecture" forum. This presentation will cover Xiaohongshu's recommendation engineering architecture, business characteristics, pain points, overall architecture, and recommendation processes, as well as how to implement graph business architecture in search promotion scenarios and improve algorithm iteration efficiency through hot deployment mechanisms.

2. "Model Training and Inference Integrated Architecture in Search Promotion Scenarios" by Jing Lun on August 26, 11:25-12:00 in the "Core Architecture" forum. This presentation will explore Xiaohongshu's machine learning engine overall architecture, large-scale distributed model training framework, model estimation platform, model optimization toolchain, inference engine, and data feature platform.

3. "Xiaohongshu's Recommendation User and Content Cold Start Practice" by Li Mu on August 27, 09:10-09:50 in the "Recommendation Cold Start" forum. This presentation will detail Xiaohongshu's practices in recommending users and content cold start, focusing on solving the problem of breaking through user circles in low-activity user behavior sparsity scenarios and content breaking through.

The summit will feature two specialized forums:

- "Engineering Architecture and Training Inference" forum, directed by Miao Tai, will discuss GPU acceleration frameworks, real-time training solutions, and training-inference integration optimization in the context of heterogeneous hardware, large-scale data, and complex model architectures.

- "Recommendation Cold Start" forum, directed by Feng Di, will address the challenges of cold start in recommendation systems, including new content, new users, and new systems, and share best practices for user cold start and content cold start across different business scenarios.

The summit will be held online from August 26-27, 2023, 09:00-12:35, with live streaming throughout. Registration is available through the provided QR codes and links.