Yusuke Takeda

Please introduce yourself.

My name is Yusuke Takeda. I joined Yahoo as a new graduate engineer in 2016. In graduate school, I was a student of a research lab focused on information retrieval. So, I wanted to apply the techniques of information retrieval in my job, and I thought Yahoo was a company where I could work on search technology. Although I had a few other options, Yahoo's status as one of the largest leading companies in Japan, dealing with large datasets, was very attractive for me.Plus, Yahoo has abundant domestic data and the opportunity I can handle it at the core of search functionality development were big factors to decide to work for Yahoo.

Since joining the company, I have been working on "Yahoo! Search." In my first year, I developed backend systems using machine learning. Then, , I participated in a project to replace the search middleware till my fifth year. Currently, I am working as a project manager (PjM) on a team dedicated to improving direct search (commonly known as Direct Display, or DD) for "Yahoo! Search."

Could you please describe your main responsibilities and the specific workflow you follow?

In "Yahoo! Search," DD (Direct Display) is a feature that directly shows the search results that users are likely looking for. For example, for a search query like "Shinagawa Prince Hotel," instead of general site links, it displays a unique answer panel with details such as accommodation plans and hotel information. We are continuously working on improving this feature to display the correct answer panels for more queries.

There are three main components involved in displaying DD: intent determination, search, and display determination, all of which utilize machine learning.

Intent Determination Component: This component determines whether the query is intended to search for a specific entity. For example, a keyword like "Yahoo" does not indicate an intent to search for a specific entity.

Search Component: This component searches for the entity to be displayed. For a query like "Shinagawa Prince Hotel," determining the entity is straightforward, but it also needs to handle variations in the query such as "Prince Hotel Shinagawa" or "Shinagawa Prince." Additionally, for queries that are not specific entity names, like "book-friendly hotel Nagoya," it needs to identify the entity the user is looking for.

Display Determination Component: This component decides whether to display the final DD and where to position it based on factors such as CTR (Click-Through Rate).

To improve these components, our team, consisting of 10-15 members works on developing systems and models to determine when to perform intent determination and how to conduct the search.

The improvement process typically starts by collecting around 1,000 queries for which we want to display DD. If only 500 of these queries can display DD, we analyze why the remaining half cannot and address the issues. This might involve extracting usable keywords from reviews or gathering new data to enhance the search engine's accuracy, tackling one problem at a time.

As a PjM (Project Manager), my role is to lead the project towards achieving its goals. This involves everything from project management, planning strategies, and conducting technical research to actual development and reviews. The DD display rate is a key service goal and directly impacts revenue, so I feel a strong sense of contribution to the company, which is very rewarding.

Example of a Daily Schedule

• 9:00 AM

  Start work. Development and code reviews for team members

• 11:00 AM

  Meeting with internal members

• 12:00 PM

  Lunch

• 1:00 PM

  1-on-1 meetings, regular project meetings, etc.

• 5:00 PM

  Development and code reviews for team members

• 7:00 PM

  End of work

Since most meetings are scheduled in the afternoon, I secure work time in the morning and evening to focus on coding adjustments and code reviews for team members. While there are many meetings, our goal of improving the DD display rate is clear, so the direction of discussions rarely deviates. Compared to other projects, making adjustments and decisions might not be as challenging.

What do you focus on when working?

In my current project, we have KPI targets, so I strive to analyze the challenges as precisely as possible and address the most significant issues first. In the past, I used to proceed with a general sense of "this area seems like it could be improved." But, by focusing on specific figures like "Doing this will improve the score by X points," it has become easier to make decisions and prioritize tasks. This approach has led to more efficient development and better results.

When we gather around 1,000 queries for improvement, I make sure not only to have team members review them but also to check as many as I can myself. I also have some reviewed by my supervisor and decision-makers. Simply looking at the results can make it difficult to grasp the challenges and their difficulty levels. So, I spend time on analysis to ensure that the process can be effectively utilized for improvements.

Could you share any interesting or challenging experiences you've had while working?

One of the most memorable projects was the middleware development project for "Yahoo! Search." Originally, we were using a system developed by Yahoo! Inc. in the United States (which has since ended its capital tie-up with Yahoo Japan), but we decided to completely replace the system and built the core search system almost from scratch.

I joined the project from the beginning, but at that time I worked for only two years and I was still quite inexperienced.. The legacy system we were replacing was quite outdated, with complex and difficult-to-understand specifications, making the transition to the new system very challenging. I spent a lot of time improving my skills, receiving code reviews from senior colleagues. It was a large-scale project that took over three years to complete, and I eventually became the technical lead for the system. When we finally released it, it was a deeply emotional moment, and we celebrated with high-fives among the team members (laughs).

In what areas do you feel your growth?

I believe I experienced significant growth through the middleware replacement project I mentioned earlier. The system specifications changed, and the impact of search results on the business was substantial. What is more, I gained business perspectives beyond development, such as "How to make decisions and drive the project forward" and "How much to consider temporary revenue declines." There were times when issues like search ad revenue not being generated that makes me feel chills running down my spine , but those challenges gave me valuable learning experiences.

Also, transitioning from an engineer to roles such as technical lead and project manager (PjM) increased the pressure to some extent, but it also broadened my scope of responsibilities. This expanded my contributions and impact, and I feel it has led to my growth.

What do you feel has been the most rewarding part of joining the company?

There are many people within the company who have extensive knowledge about search technology, making it a great environment for learning as an engineer. And, there is a significant advantage as we can work on improvements using large amounts of data. In machine learning, the quantity and accuracy of data are crucial, and having access to large volumes of high-quality data makes it easier to develop precise models, which is ideal for an engineer.

In terms of work style, the shift to full remote work during the COVID-19 pandemic was another positive aspect. It allowed me to work from anywhere in Japan, so I moved from Tokyo to Osaka after getting married. The full-flex system with no core hours also makes it easier to manage my own time and allows me to maintain a good work-life balance.

What are your expectations and things you are looking forward to at LY Corp in the future?

Since LINE and Yahoo were separate companies, I believe there were differences in the speed of decision-making and other aspects. As LINE and Yahoo merge, I am personally involved in how we can strike a balance and bring about new changes. I am also looking forward to the cultural integration.

On top of that, becoming one company will increase the amount of data we can handle, which will undoubtedly enhance the performance of our search engine. LINE has a service called "LINE PLACE" where users can search for spots and write reviews. For example, if there is a review on "LINE PLACE" that says "This restaurant has a happy hour," through data integration, we will be able to display this restaurant when someone searches for "happy hour" on "Yahoo! Search." This is just one example, but with more data, it will be exciting to see how we can improve our services and solve user issues more effectively.

What are your future goals?

At my core, I always want to contribute to the service and help solve user problems. My contributions themselves are primarily through data-driven improvements and technical approaches, so my focus on technology will not change. However, I am not overly focused on data itself; my mail goal is to contribute to the service, and technology is just a tool to help achieve that.

As a PjM (Project Manager), my objective is to quickly organize a team that can deliver results efficiently while maximizing contributions to the company. I always want to help team members gain various experiences and encourage them to take on new challenges. I believe this will ultimately lead to the company's growth.

Please tell us your final message.

In the field of information retrieval technology, there is a growing movement to tackle challenges that used to be difficult with traditional systems by using generative AI, such as ChatGPT, to extract information from reviews. While there are still many uncertainties, I am very excited about the future evolution of this technology.

What is more, as the company has expanded, we currently see a very valuable point where we are creating new things. There have always been many talented people, diverse roles, and large amounts of data in each company, but becoming one company allows us to work with even more professionals, handle a wider variety of tasks, and solve problems using even more data. The number of users we can impact will also increase dramatically, which make it an incredibly exciting environment for an engineer dealing with information technology. Being able to work in such an environment is the most attractive aspect, and personally, I am very much looking forward to working with a variety of people in the future.

• Engineer
• Mid-Career