Yujie Liu - Driving Impact Through Data

Many of our employees here at Wish make an extraordinary impact that helps to drive the company's revenue and inspire others.

I sat down with our Senior Data Scientist, Yujie Liu, to explore and understand the development of his RSO Model and how it generated over one hundred million in transaction value for Wish.

Before we dive in, let’s learn about how Yujie Liu started working for Wish. He started towards the end of March 2018 as a data scientist intern. In this position, he discovered how to utilize his skills to move product direction and overcome obstacles for Wish’s data analytic systems while developing expertise as a machine learning engineer.

The RSO Model

During his time at Wish, Yujie has worked in merchant fraud, user profiling, and fraud transaction detection. He developed the RSO Model and the relevant modeling infrastructure, which has helped data scientists launch risk-related machine learning models quickly. With the help of the Buyer Risk Team under Wenbo Zhang and another data scientist, Xinyu Liu, Yujie was given the freedom to work with payment engineers.

Read the full interview below to learn more about how Yujie has improved model development and management efficiency through his RSO Model here at Wish.

1. What first attracted you to Wish as an Intern?

Yujie:  "I enjoyed the diverse community and the many types of people working at Wish. The teams are energetic and aggressive, and the managers permit employees the freedom to try out many different things. Wish has been moving extraordinarily fast since I joined in 2018, and my favorite part is how everybody is open to making big changes to improve this company in any way possible."

   2. What’s your role at Wish?

Yujie: "I am part of the Data Scientist team and hold the title of Senior Data Scientist. My team and I are in charge of analyzing Wish’s data to identify different patterns and trends within the company. We try to figure out what we need to do to positively move product direction while following the nature of science. The project I am most proud of is the RSO model and the relevant modeling infrastructure I developed."

   3. What is the RSO model and how has it made an impact at Wish? 

Yujie: "The RSO model is a transaction model to classify transactions to be fraud and non-fraud. I developed this project in March 2020 and it has already generated over one hundred million in transaction value toward Wish since then."

   4. What achievements have you accomplished at Wish thus far?

Yujie: "My RSO model has brought about over one hundred million in transaction value to Wish since March 2020. To finish the development of this model, I have completed tasks in different areas as preparations: I worked with the payment engineering team to develop the signal aggregator collections, which aggregate the information to user-level and enable the data scientists to fetch such features in real-time in the production system. To manage the scores from the RSO model, I developed the rules engine POC to have a highly customized strategy to treat different segments. To improve the efficiency of the operation team who are helping provide feedback for the RSO model, I proposed the project of an enhanced manual review queue. We redesigned the interface and offered various management strategies to both external and internal teams. Lastly, to avoid duplicate work, I developed the feature factory to make it available for other data scientists to easily apply the existing features and focus on developing new features only. This feature factory was leveraged in experiment ground, which was finished by data engineers and myself and has provided an easy approach to conduct feasibility studies of projects related to models and rules. To conclude, all the preparation work was not only for the development of the RSO model but also for improving the efficiency of model development and management."

   5. What advice would you give future data scientists at Wish?

Yujie: "A word of advice to the next generation of data scientists: although data is helpful, it can be very confusing. Assume many things and get first-hand info because it will avoid misunderstandings and improve team efficiency. While balancing science, also follow the nature of business. If some results are not as effective, do not play any tricks! Lastly: always be open to learning new things! Do NOT wait for someone to resolve things for you. Be adaptable and learn how to pick up new things on your own."

Here at Wish all of our employees have equal opportunities to make an impact. Are you thinking about joining the Wish team? Check out all of our available positions below.