"Recursion" Episode 9 | I'm doing research at Tencent

Why are we called "recursion "

"Recursion" (recursion) is an algorithm that is widely used in programming languages. It has two major characteristics, one is to call itself, and the other is to simplify the complex. Isn't that the case with the excellent technical people among us? They lead by example, use practical actions to influence those around them, standardize the process, improve efficiency, and benefit the entire team. This is the original intention of our "recursion" column, to record the extraordinary of ordinary Tencent technical people.

In Tencent, there is such a group of young technical "craftsmen". They focus on scientific research and pursue the ultimate. Looking up at the starry sky of cutting-edge technology, but also stepping on the field of technology application. Today, let's listen to their stories.

Chen Xiangyu , graduated from the University of Tokyo with a Ph.D., in the TEG Robotics X laboratory of Tencent, mainly engaged in the design of robot software and hardware system architecture and the research of robot mobility.

Qian Yuqiu , a Ph.D. graduated from the University of Hong Kong, is responsible for the algorithm research and platform engineering construction work in the game recommendation scene in the Tencent IEG Value-added Service Department.

Danyl , Ph.D. graduated from the LAMDA Institute of Nanjing University, engaged in computer vision application research in Tencent CSIG Youtu Lab.

Deep plowing: on the road to scientific research, the achievements are worthy of hard work

Play the game, explore the frontier

Chen Xiangyu, who focuses on robotics research , participated in the Robot Challenge organized by DARPA (US Department of Defense Advanced Research Projects Agency) when he was a Ph.D. student at the University of Tokyo. It is considered to be the world's highest-spec robot competition, and it can be called the Olympics in the robot world.

Chen Xiangyu has benefited a lot from participating in high-level competitions. He feels that these competitions " capture the most cutting-edge in the robotics industry and some common problems faced by all researchers, and let me know how to move forward on this road. Direction ".

After coming to Tencent, Chen Xiangyu worked in the Robotics X Robotics Lab to research on the mobility of footed robots, enabling them to have a perception system, interact with people, and give them very dexterous movement capabilities.

Real knowledge from practice

Qian Yuqiu , who is also a Ph.D. graduated from the University of Hong Kong, is responsible for the algorithm research and platform engineering of game recommendation scenarios in the Tencent Interactive Entertainment Group. At that time, she saw that the company's online algorithms seldom used sequence recommendation, so she led the interns to make a new sequence recommendation algorithm, and published in "Lego Infinity", "Tianya Mingyue Knife", "Yulong Zaitian Mobile Games", etc. The game went online and finally achieved good results.

When you first joined Tencent, you have to undertake the construction of a system sub-module. Yu Qiu was particularly unfounded at the beginning, because theoretical knowledge in books and practical training in the industry are two concepts. But she did not retreat, instead, she faced difficulties, learning while doing and asking while learning. In the end, the entire system ran through and was unanimously approved by the team.

Technology empowerment, fast and accurate

The computer vision R&D team of Tencent Youtu Lab where Danyl works is developing an industrial defect quality inspection project. In order to help the factory improve the efficiency and accuracy of quality inspection, they developed an automated defect quality inspection system, which can process millions of defect images every day through AI image algorithm identification, helping the factory save about 50-70% of manpower. The system has been online for one year, helping companies save more than 10 million in costs.

Joining WTO: Stepping out of the comfort zone and entering the industry

Several doctoral students also have different opinions on the difference between doing research in universities and companies.

Do more influential things

Chen Xiangyu: When doing scientific research in school, many times it is based on the research team of several doctoral and master students led by the supervisor. I came to Tencent and found that there are many experts in different fields and different directions. Cooperating with them can make one thing bigger and exert greater influence.

Proactively find and define problems

Qian Yuqiu: In schools, many times, after conducting research and algorithm iteration for existing problems, the effect of offline standard data sets is improved. In the company, many problems are not defined, and you need to discover and explore by yourself. At the same time, the research results need to be improved not only in the standard data set , but also in practical applications and in some indicators that the company pays attention to.

Cross-team collaboration

Danyl: In school research, I usually research a problem with my supervisor and fellow seniors . In the company, it is faced with a series of customer problems and needs to provide a complete set of solutions. This means that a project is difficult to complete with the power of a few people and requires cross-team and cross-department cooperation.

‍

Original intention: N reasons to join Tencent

Research fit

Chen Xiangyu: My mentor’s friend and well-known scholar, Dr. Zhang Zhengyou, joined Tencent in 2018 and established the Robotics X Robotics Laboratory, which is very consistent with my research direction. I also want to challenge to join the industry to understand what the company is researching.

Explore and share

Qian Yuqiu: Here you can freely explore the problems and solutions of interest . Secondly, there are many like-minded friends around, and everyone is keen on logical thinking and technical exploration. In addition to academic discussions together, we also have many opportunities for external sharing.

"Research results can only produce value"

Danyl: Tencent Youtu Lab is one of the top machine learning research teams in China , especially in the field of computer vision research leading the country. In addition, Youtu Labs pays attention to the principle of "research results can only produce value" , which is in line with my original intention of computer vision research: using technology to improve our lives.

After reading the stories of three young technical "craftsmen", what do you want to say to them? Please leave a message and let me know! We will select three lucky readers and give a beautiful gift.