E-mail / Google scholar / DBLP / 中文版
Formerly a Senior Algorithm Expert at Alibaba (Alias: Da Qing). Both his Bachelor’s and Master’s degrees were obtained from the Department of Computer Science and Technology at Nanjing University, supervised by Professor Zhi-Hua Zhou, focusing on machine learning research. In early 2015, he withdrew from his Ph.D. program during the second year to join Alibaba’s Search and Recommendation Division, working on search and recommendation algorithms.
During his tenure at Alibaba, in addition to his core responsibilities, he was involved in training new algorithm talents. He was a co-author of the 2019 group newcomer course “Introduction to Search, Recommendation, and Advertising Algorithm Systems” and served as the main lecturer for the Hangzhou session. He also served as the head teacher for the Search and Recommendation Division’s algorithm newcomer class in 2020. Externally, he has served as a reviewer for top-tier journals and conferences such as TNNLS, AAAI, IJCAI, and ICML.
He joined ByteDance in October 2022, focusing on e-commerce algorithms.
Education
Bachelor’s Degree: Nanjing University, Department of Computer Science and Technology, September 2006 – June 2010
-
First Place (National First Prize) in the National Undergraduate Mathematical Modeling Competition, first author, 2008
-
Ministry of Education – Sun Microsystems University Student Innovation Experimental Plan Outstanding Project, first author, 2009
-
Nanjing University Undergraduate Excellent Thesis (Design) First Prize, 2010
-
Recommended to continue graduate studies in the same department (Rank 1st)
Master’s Degree: Nanjing University, Department of Computer Science and Technology, Machine Learning and Data Mining Laboratory (LAMDA), September 2010 – June 2013
-
PAKDD Data Mining Competition Public Division Champion, fifth author, 2012
-
First “China Cloud · Mobile Internet Innovation Award,” One First Prize, One Second Prize, Two Third Prizes, first author, 2013.
-
Master’s Thesis: “Research on Reinforcement Learning Methods Based on Direct Policy Search”
PhD (Incomplete): Nanjing University, Department of Computer Science and Technology, Machine Learning and Data Mining Laboratory (LAMDA), September 2013 – January 2015, Discontinued
-
During the first year of PhD, published one CCF A-class paper [2006-2] and one CCF B-class paper [2006-3] as first author. Mid-program evaluation ranked excellent (Rank 1st).
-
Withdrew from studies at the beginning of the second year for personal reasons to enter industry.
Professional Experience
January 2015 – June 2016: Senior Algorithm Engineer, Search Business Unit, Alibaba
-
Engaged in Taobao search algorithm development work
-
Through collaboration with the engineering team, successfully deployed the first version of a large-scale online learning model based on parameter server, promoting the widespread application of online learning in Taobao search. The related work “Real-time Personalization for Taobao Search Based on Online Matrix Factorization” won the 2015 Group Top Ten Algorithm Award
-
First introduced the technology of optimal allocation combined with PID online adjustment to Singles’ Day red envelope distribution. This technology has become the foundational general solution for group-wide rights distribution and traffic control, and resulted in the patent “A Method for Distributing Red Envelopes via Keywords to Sell Traffic”
-
Recipient of the 2015 Search Business Unit New Employee Award, Double Eleven Crazy Search Person Award, and Technology Excellence Award team member
-
Multiple internal group algorithm competition awards: Runner-up in Sina Weibo Interaction Prediction Competition; Runner-up in Resume Intelligent Scoring Competition; Cainiao Demand Forecasting & Warehouse Planning Winning Award (rank 4)
July 2016 – December 2017: Algorithm Expert, Search Business Unit, Alibaba
-
First introduced reinforcement learning into e-commerce search ranking, with online application during the 2016 Double Eleven shopping festival, covered by multiple media outlets. Machine Heart Report
-
Designed and implemented the AI4B-RL reinforcement learning framework for internal search, applied reinforcement learning across multiple business lines, including search ranking [2018-2], product display learning [2018-3], engine performance optimization [2021/2022-6], seller stratification control, traffic control, virtual Taobao [2019-3], and other projects. Related work has been published in KDD, AAAI, and ECML.
-
Organized all related algorithm teams across the group and published an e-book “Technical Evolution and Business Innovation of Reinforcement Learning at Alibaba,” covering broad domains including search, recommendation, advertising, logistics, and intelligent customer service. The book was subsequently published in physical form by Electronic Industry Press.
January 2018 – July 2020: Senior Algorithm Expert, AI International Business Unit, Alibaba
-
Led a team to participate in the OpenAI-hosted reinforcement learning algorithm competition, winning the global championship, as first author. OpenAI Official News | Alibaba Technical Report
-
Starting August 2018, assumed the role of AliExpress Search Algorithm Lead, building a team and driving technology upgrades tailored to international business characteristics. The seed team of 2-3 people has grown into a 20+ algorithm team, covering four sub-directions: search relevance, search efficiency, search discovery products, and traffic control. Over two consecutive fiscal years, search conversion rates improved by 30% and 20% respectively over the baseline that had been optimized for 8 years, with significant improvements in search relevance, search discovery products, and related business metrics.
-
Addressing multilingual challenges in international business, began researching cross-lingual vector models around the same time Facebook proposed XLM, currently covering English, Russian, Spanish, French, and Portuguese, fully deployed for semantic vector recall in search [2021/2022-3]. To address difficulties international sellers face in title writing, established relationships between images and keywords to assist in title completion. This work was organized and published at PAKDD [2021/2022-2].
-
Given that single models struggle to capture behavioral differences across global users, upgraded the classical MOE structure to a hierarchical MOE structure, using scenario sub-networks and scenario gradient isolation to ensure final fusion effectiveness. This work has been fully deployed online and was accepted at CIKM’20 [2020-2].
-
For the re-ranking scenario, to capture user behavior preferences regarding overall product arrangement, proposed a learning-to-rank framework based on an evaluator-generator architecture that can directly generate product arrangements to optimize full-page user behavior. This work has been deployed online and was published in TKDE [2023-1], with related media coverage
August 2020 – September 2021: Senior Algorithm Expert, AE Technology Department, Alibaba
-
Designed and productized the AliExpress traffic control system for efficient integration of commercial and efficiency strategies. Related work includes online matching [2021/2022-7] published at AAAI 2021, with remaining components currently under review [2021/2022-10].
-
Currently serves as AliExpress Search Algorithm Lead and additionally holds the role of AliExpress Algorithm Architect, responsible for cross-functional capability building for AliExpress’s multiple business algorithm lines including search, recommendation, advertising, and user growth.
October 2022 – May 2024: Head of Content E-Commerce Recommendation Algorithm, ByteDance (Douyin)
-
Led a team on content distribution e-commerce content types on Douyin, including but not limited to shopping cart videos, store anchor videos, e-commerce video avatars, e-commerce live streams, and non-linked e-commerce intent videos. Responsible for improving matching efficiency between users and e-commerce content while meeting user content interaction constraints, optimizing pre-purchase and post-purchase user experience, fostering effective interaction with e-commerce primary scenarios (such as shopping malls and store windows), and enhancing overall e-commerce penetration and transaction scale.
-
Key projects include: e-commerce traffic framework establishment, e-commerce interest capture, balance between content experience and e-commerce efficiency, consumer growth technology, and e-commerce traffic control systems.
June 2024 – Present: Head of TikTok E-Commerce Search Algorithm, ByteDance
-
Leading a team on TikTok e-commerce search algorithm optimization, covering all consumer-facing e-commerce search scenarios including TikTok general search e-commerce, shopping mall search, and image search.
-
Key projects include: Query understanding and relevance optimization, full-link efficiency optimization for search including retrieval/coarse ranking/fine ranking/mixed ranking, and e-commerce experience and ecosystem cold-start optimization.
Publications
BOOK
Da Qing, Zeng Anxiang (Editors). Reinforcement Learning in Practice: Technical Evolution and Business Innovation of Reinforcement Learning at Alibaba. Electronic Industry Press, 2018.
OPEN SOURCE RECOMMENDATION SYSTEM ENVIRONMENT
VirtualTaobao [AAAI’19], Github: This project provides a VirtualTaobao simulator trained on actual Taobao data, addressing the problem of evaluating learning strategies in traditional recommendation system models using biased offline data. At Taobao, when a customer inputs a search query, the recommendation system returns a series of products based on the query and customer information. The system aims to return a product list with high click probability. Through the VirtualTaobao simulator, users can access a “real-time” environment similar to the actual Taobao environment. Each session generates a virtual customer who initiates a search query, and the recommendation system must return a series of products. The virtual customer decides whether to click on products in the returned list, similar to real customers. The training procedure of VirtualTaobao is described in the original paper:
- Jing-Cheng Shi, Yang Yu, Qing Da, Shi-Yong Chen, An-Xiang Zeng. Virtual-Taobao: Virtualizing real-world online retail environment for reinforcement learning. In: Proceedings of the 33rd AAAI Conference on Artificial Intelligence (AAAI’19), Honolulu, HI, 2019. PDF
2023
2021/2022
2020
2019
2018