Xiaodong He

Overview

Xiaodong He is vice president at JD.com and deputy managing director of JD AI Research, according to JD's corporate blog. He is also an IEEE Fellow; IEEE Fellow is a recognition that IEEE gives to selected members. At JD, his work is associated with conversational AI: systems that let people ask questions and receive answers in natural language. Such systems combine language understanding, dialogue management, and answer generation, and in a company setting they can handle customer questions at scale. A JD corporate blog post dated 21 April 2021, written by Ella Kidron from a conversation held on 2 April, describes this work and the team around it. ^[1]

Role and team

The same blog says that the conversational AI team he works with grew from 10 people to 300. The change from 10 to 300 is a large expansion: it marks a move from a small group to a much larger organization. JD had more than 471 million active customers as of December 2020, which gives a sense of the user base that such systems can serve. For a company of that size, even a small share of questions can mean a large volume of interactions, so the scale of the team and the scale of the customer base are connected parts of the story. ^[1]

Research and competition

JD's AI research team was ranked No. 1 at Stanford University's QUAC question-answering-in-context competition, according to the blog. QUAC is a question-answering-in-context task: a system must answer questions using a supplied context rather than only matching keywords. That makes the task a test of finding the relevant part of a passage and producing a short answer from it. The No. 1 ranking is a public competition result for the team. ^[1]

Safety and emotion detection

The blog describes JD's Fine Granite emotion detection system noticing a user's query about sleeping pills and passing it to JD's Life Channel suicide hotline team. This is an example of a conversational system doing more than returning an answer: it detects a potentially urgent signal in text and routes the case to a dedicated support team. Emotion detection adds a classification step to the dialogue flow; the system estimates whether a message carries a signal that needs special handling, and then a separate team can take over. ^[1]

Public-service hotline

JD's chatbot technology was also applied to the 12345 resident service hotline of Datong in Shanxi. According to the blog, daily calls rose 31.7% and the pickup rate was 100%. The case places conversational AI in a public resident-service setting, where call volume and pickup rate are practical measures. A 100% pickup rate means every call was answered, while the 31.7% rise in daily calls indicates more volume; together the two figures describe capacity and access. ^[1]

Why it matters

For a person entry, these facts place He at the intersection of corporate AI research and deployed conversational systems. His titles and IEEE Fellow status describe his professional standing. The team growth, the QUAC result, the emotion-detection routing example, and the 12345 hotline application describe a body of work associated with him. Taken together, they show a path from research and competition to customer service and public hotlines. ^[1]