ReachOne

Visionalのイマとこれから

They've even done thesis presentations in Alaska! BizReach's AI Group — advancing our products synergistically

Hi. This is Tomisato from the ReachOne Editorial Team.

The AI Group was founded in 2016 as a team of specialists. The team applies knowledge they have gathered from conducting research and development on the latest technologies involving machine learning and data science to BizReach's several projects.

I am glad to have Shinsuke Sugaya, one of the founding members and the current manager of the AI Group, to share with us the role AI Group plays in BizReach and the future prospects they have for it.

f:id:bizreach:20191122125109j:plain

Shinsuke Sugaya
Shinsuke has worked as a Product Director, etc. for major foreign software development companies. In 2014, he joined BizReach and was involved in research and development for Stanby, and in 2016 he became the manager of the AI Group. Also, he is a committer to several of Apache's open-source products and is one of the developers of the full-text search server, Fess. In recent years, Shinsuke is known for his work on machine learning research and is a committer to Apache's machine learning server, PredictionIO.

Highly-Specialized Members Assemble Around the Globe

—What led you to join BizReach?

I have known BizReach's CEO, Soichiro Minami, and CTO, Shin Takeuchi, from around the time they founded the company. In 2014, Shin invited me to "create a cutting-edge search engine" together. Through his words I felt his passion and was moved by his high aspirations. At the same time, it was a chance for me to take advantage of my accumulated knowledge of machine learning. That was why I decided to join the founding members of BizReach's job postings search engine, Stanby.

I focused on improving the quality of Stanby's search results. This involved continuous quality improvement with the engineers as we aimed to surpass Google's search quality in the HR category. As a result, in February of 2016, Stanby became the search engine provider for Yahoo! Job Search (formerly Yahoo! Job Postings).

—So Yahoo! took notice of Stanby's impressive technological prowess! How exactly was AI Group created after that?

Originally, each department had members in charge of data science and machine learning. These departments had developed their own AI technologies individually.

However, it would be inefficient to have multiple teams developing separately when most products required similar AI technologies.
Also, we plan to consolidate talent sources who are knowledgeable in AI and provide support to leverage their expertise. So, a team specializing in AI was established to fulfill these requirements.

—Around how many members did the AI Group have when it was formed?

The AI Group had 5 members, including me, when it was formed in 2016. Now we have 12 members and we're expected to grow to 16 by next year. 3 of us have PhDs—5, if we include our recruits for next year.

We are a highly specialized global team with members not only from Japan, but also from America, India, China, and Vietnam. Half of our members are foreigners, so we are the most diverse department in BizReach. Although a bit of help is still needed for some of our members who are not yet fluent in Japanese, I conduct 1 on 1 meetings with them to uncover whatever issues they may encounter, which we would then address as a team.

—It seems that this team has gathered a multitude of capable people from around the world. So, what is it about BizReach`s AI Group that makes it attractive to talented and accomplished professionals with advanced technological know-how and abilities and draws them into being a part of it?

They all have different reasons. Many of them in the due course realized that at BizReach, we create and maintain our projects. This gives them a great deal of control and broadens their horizon as they could handle a variety of data, especially with regards to HR. I guess this is what makes BizReach appealing to them.

We have a lot of rare datasets that you wouldn't find anywhere else—BizReach's executive-class talent database, Stanby's enormous quantity of job posting data, etc. Analyzing such data is rewarding, albeit challenging. That's one of the reasons why most of our members have a particular interest and are highly knowledgeable in Natural Language Processing. *1

Also, an outsourced data scientist would have to jump hurdles before they can access and analyze a client's database. However, that sort of inconvenience does not happen in BizReach—we can work together with the products' engineers and develop technologies without going through the hassle. I would say another huge point of interest is that BizReach existing products and startup projects are growing continuously, so this allows us to apply the technologies we develop to various services. f:id:bizreach:20191119103904j:plain

Presenting Their Research at an International Conference!

—Tell me about the interactions AI Group has with the other departments.

We have regular meetings with representatives from each department. During these meetings, we ask them what they want to do with AI technology, we tell them what we can offer, and we decide on what we are going to do specifically. Each product is in a different phase and wants to do different things, so it's essential to communicate thoroughly and understand what each of them needs.

—How is AI technology being used in each service specifically?

Each service is different. For example, BizReach and CareerTrek make use of machine learning for job recommendations. Job postings searching engine Stanby, use previous data to calculate and display an estimated annual salary for job postings that do not provide that information. Business media BizHint, segments its users and visualizes the correlation between a user's interests and job title.

Also, BizReach Campus, a network for OB/OG meetings, conducts research on the correlation between OB/OG meetings and getting a job offer from the company.

—The AI Group has deeply committed to research and development and even done thesis presentations, right?

Yes. It's not part of their work requirement, but we have members who join competitions in machine learning and data science or are involved in research projects developing the latest in AI and machine learning technology.

"Recently, we were in Alaska to present in a workshop on Knowledge Discovery and Data Mining 2019 (KDD 2019), an international conference on data mining held by the Association for Computing Machinery (ACM).

● Research Topic
Extrapolation of an interviewer/interviewee' skills

Based on data sets derived from interviews, the interviewer's strictness/kindness ratings, and the interviewees' skills are extrapolated and visualized.

Other than that, we have also presented for Japanese associations such as The Association for Natural Language Processing (ANLP) and The Japanese Society for Artificial Intelligence (JSAI).

Having many members with highly specialized academic knowledge is one of the AI Group's strengths. I intend for the team to continue research on new technologies worthy of being presented in international conferences and remain at the forefront of AI technology. The AI Group is set to continue to actively join more international conferences in the future.

f:id:bizreach:20191119103933j:plain

Making Products Edgier with AI

ーWhat kind of team is the AI Group striving to become?

BizReach's startup projects continue to increase in number. Because the number of AI Group members has been growing, the scopes of such projects which we are able to assist are also broadened. We want to help various projects grow by strengthening coordination among the business functions.

For example, we are conducting research to improve yamory with AI technology. yamory is an OSS vulnerability management tool and as such, gathering the latest information on vulnerabilities is crucial to the service. There is an abundance of information on the Internet, using AI technology we have ongoing research that would enable us to extract only the information related to security and sort them into data that can be utilized by yamory. The ability to automatically identify meaningful information will prove to be another one of yamory's assets.

Whenever a new service is released, competition is bound to follow soon after. Because of this, it is essential to make your product stand out from the crowd. The AI Group aims to harness the power of AI in order to develop functionalities that would differentiate BizReach's products from its competitors.

Furthermore, we also pursue new research in preparation for future prospects. For example, we are researching OCR *2technologies to digitize forms our customers have filled out by hand and developing chatbots to automate FAQs which reduce redundancy in our internal operations. By conducting in-house R&D and applying them within the company beforehand, it allows us to apply AI to our products more efficiently.

Thank you. I'm looking forward to seeing how each product will evolve further through the use of our technologies.

About the Writer

Shimpei Tomisato


Shimpei joined BizReach in 2015 after having worked in online advertising. As part of the product planning team, he was in charge of the service strategies and CRM for B2B Marketing and BizReach. Since transferring to the HR/Recruitment Marketing Group in August 2018, he has been in charge of the company's recruitment branding and owned media.

*1:Natural Language Processing: Languages humans use for daily communication are called "natural language". On the other hand, programming languages are known as "artificial language". Natural Language Processing refers to the technology used to understand, interpret and manipulate natural languages as used by man.

*2:OCR: A technology that is used to scan hand-written or printed characters and convert them into digital characters recognizable to computers.