Letters from Leaders

I hope you’ll go one step further: decide which problems are worth solving, and what the world should look like once you’ve solved them.

Shiyingxue Li

Senior Vice President, Data

Indeed Recruit Technologies Co., Ltd.

Shiyingxue Li laughing during a conversation in a Recruit office

I’ll start with a confession. When I met Recruit back in 2013, I wasn’t going to apply for a job at all.

Well, I did take a look at engineering jobs out there, and they all seemed to work the same way: someone else decides what to build, and you build it. I had spent my whole life deciding for myself what to do. Research looked like the last place on earth that would let me keep doing that.

But research had two little difficulties.

First, research needs data, but data doesn’t always come easily in academia. I was studying sentiment analysis on reviews, and the first step is to figure out a way to get data. Picture trying to study the ocean with a bucket.

Second, and this one mattered more. After I finally got a paper accepted, I asked myself: “Okay. Suppose this is the best algorithm in the world. When will it matter to anyone but me?” The answer was probably a few years at best, if at all.

So when I sat down for my interview with Recruit, I didn’t ask polite questions. I said, “I want data, and money to buy enough machines. Also real users for tests. And while we are at it, I want to be able to decide what to build. Is that something you can do at Recruit?”

The answer was short: yes.

To be completely honest with you, I didn’t quite believe it. But I was curious enough to move to a country whose writing I couldn’t read. I didn’t know a single hiragana. I told my interviewer, “I’ve learned English, surely I can learn Japanese too.” Then the adventure started.

Like how a good adventure story should start, in my first year, I was what you might call “a problem.” My manager tried to train me by giving me a textbook. I said, “But I don’t have any project that actually uses stuff in that book. Maybe another time.” That was the start of years of arguments. There was one 30-min performance review that went on for several more hours because I had different opinions on what should be considered great (and yes, the scale was Exceeds Expectations.)

Here’s what I eventually figured out.

In school, if people don’t understand your brilliant idea, that’s their problem. In the real world, it’s yours.

Being right and being heard are two different skills, and I only had the first one. I learned the second at SUUMO, our real estate site. Ten years ago, if you searched for an apartment near downtown Tokyo, the results would be ordered by price, so the top one would ask you to commute from a different prefecture for more than an hour. Technically correct. Not what the user asked for in most cases. I wanted to add a personalized ranker so some people could actually see what was nearby. The search product team said you can’t run a model in real time over millions of listings. They had a point.

So instead of winning the argument, I sat down with the team to learn how they have built the current search engine over the years. We discussed what part absolutely shouldn’t be touched, and what heavy-lifting could actually be worth it.

Months later, the new ranking model significantly boosted business KPI, people in dire need of shorter commute time got to see those homes. The work from that time became a published paper along with a few company rewards. More importantly, it became our team’s work, our team’s idea, not just mine. Everyone would agree that users’ preferences matter the most.

Today I lead the data organization at Indeed Recruit Technologies. Our mission is simple to say: no one should miss an opportunity just because it was hard to find. We want a world where you don’t need to settle for a job that’s 60% match when a 100% is out there. Along that line, something much bigger is coming. AI is going to change how every one of us works. We do not fully understand yet what work should look like on the other side of that change. For the next five years, that’s what I want to work on, with people like you: a world where people grow with AI instead of being left behind by it. Where people and companies find each other not through a list of skills, but through the things that are hard to write down: a company’s DNA, a person’s real potential and strengths.

You are clearly very good at solving problems. I hope you’ll go one step further: decide which problems are worth solving, and what the world should look like once you’ve solved them. Out here, outperforming the benchmark, if there is any available, is the start, not the end. Before you can optimize anything, you have to decide what “good” even means for millions of people looking for something important, like a home, or career.

If you’re curious whether you and your research could change how millions of people live and work, come and find out.

Shiyingxue Li

Profile

Current

Senior Vice President, Data, Indeed Recruit Technologies Co., Ltd.

2024

Became Vice President, leading data science and engineering for HR products.

2023

Moved from housing to the HR domain as head of HR data solutions.

2019

Became a manager, leading machine learning engineering for SUUMO machine learning team.

2013

Joined Recruit as a global new graduate through the WORK IN JAPAN program. Built SUUMO’s real-time recommendation system and personalized search ranking.

As a student

B.S. in Electronics Engineering and Computer Science, Peking University. Researched data mining and natural language processing.