Interview with Dr. Wang Zhongyuan: AI Research, Knowledge Graphs, and Career Journey

On January 21, 2019, MIT Technology Review announced its 2018 "35 Innovators Under 35" list for China, naming Wang Zhongyuan, head of Meituan‑Dianping AI Platform’s NLP Center, as a "Visionary".

Wang originally entered Renmin University to study International Economics and Trade, but was transferred to the Computer Science department, where he discovered a passion for computing. Influenced by professors such as Meng Xiaofeng, he learned the importance of solid fundamentals (data structures, operating systems, compilers, computer architecture) for rapid skill acquisition.

During his undergraduate years he joined Meng’s WAMDM lab, contributing to projects like the first native XML database OrientX and a Deep Web data‑integration study. His paper on entity recognition in Deep Web integration was accepted at the NDBC 2006 conference, and he received the SIGMOD 2007 Undergraduate Scholarship (one of only seven worldwide).

After completing his master’s degree, Wang turned down offers from Baidu, Tencent, and IBM to wait for a position at Microsoft Research Asia, where he eventually received the first offer among his cohort. At Microsoft he worked on knowledge‑graph and dialogue‑robot projects, gaining exposure to world‑class researchers.

In 2010, Bill Gates visited the lab, and Wang had the chance to demo his early knowledge‑graph work, receiving strong positive feedback that reinforced his research direction.

Wang spent over six years at Microsoft Research Asia, publishing more than 30 papers at top conferences (VLDB, ICDE, IJCAI, CIKM) and winning the ICDE 2015 Best Paper Award.

He later joined Facebook as a Research Scientist, leading the product‑level entity‑linking service for the social network, improving performance by roughly 80% within six months. The technology now powers Facebook’s search, recommendation, advertising, and intelligent‑assistant systems.

In 2018 Wang returned to China, choosing Meituan over other offers (including Alibaba, Baidu, and several internet startups) because he believed Meituan’s massive, diverse data ecosystem offered the best platform for AI impact.

At Meituan he built the AI Platform’s NLP Center and later the Meituan Brain – a large‑scale knowledge graph covering 23 concepts, 1.8 billion entities, and 600 billion triples – serving search, SaaS, finance, delivery, and customer‑service applications.

Wang emphasizes that solid fundamentals, data integration, and a robust technical stack are essential for successful NLP deployment in industry. He also stresses the importance of patience, belief in one’s values, and continuous learning for research‑oriented talent in corporate settings.

Regarding knowledge‑graph research, Wang notes that while many algorithms are mature in academia, scaling them to billions of entities poses engineering challenges that require new algorithmic designs and system architectures.

He advises aspiring AI professionals to focus on core technologies, understand business scenarios, and build a strong theoretical foundation (probability, analysis) before diving into programming languages.