Welcome back to PayPay Professionals, a snapshot into the incredible professionalism of the pros that drive the PayPay group.
Today we sat down with Katsunobu-san, a data scientist spearheading our company-wide AI utilization movement, to hear about recent initiatives and the meaning and principles he holds in his work.
Katsunobu Shoji
Data Strategy Promotion Team, Data Strategy Department, Legal & Risk Group
After graduating university, Katsunobu-san experienced system development at a major IT vendor in the fields of finance, transportation, and electricity. He thereafter went freelance as a system engineer, working on projects such as financial derivatives, drone autopilot, and map rendering applications. Then, intrigued by deep learning, he took on AI-related proofs of concept at major IT vendors, and talent analysis in the HR industry, finally joining PayPay in November 2022.
Why Choose PayPay in the Dawn of the AI Era?
What drew you to data science and AI, Katsunobu-san?
Fresh out of university, I wanted to get a job at a robotics company, a wish that didn’t work out at the time. Then, when terms like “big data” and “deep learning” started to trend, I saw my chance to get on that path. Robots just seemed so cool, so interesting ― ChatGPT, for example, is a robot with a command of language.
Then, I built up experience through competitions like Kaggle’s while working on projects such as AI-related POCs at a major IT vendor and talent analyses in the HR industry.
Why did you decide to join PayPay?
While continuing as a freelancer was one option, I began to think that placing myself in a company with its own data was another. PayPay, with its huge and highly reliable data infrastructure, really spoke to that thought. I figured I would be able to perform analyses toward all kinds of objectives, given tens of thousands of users’ data would be there for even niche segments.
One more point was the quick correspondence during the hiring process. Communication went more smoothly than with any other company, which made me realize how truly fast PayPay works.
Ideas vs. Criminals: Patenting an Analysis Tool to Fight Financial Crime
First of all, what is your mission?
My mission is to maintain data governance in the Data Strategy Department of the Legal & Risk Group while also promoting company-wide data utilization, which I propel from the bottom up; I gather info on demand for data analysis and AI from teams across the company then provide individual resolutions, and company-wide solutions in parallel.
So you work across teams. What specific initiatives do you work on?
I work with breadth: building network analysis tools for money laundering and account takeover analysis by the Anti-Financial Crime Department, creating data for behavioral age index analysis, or utilizing large language models (LLMs) in my local environment.
I also share data analysis-related know-how in an open Slack channel to increase our company’s overall level of skill and knowledge. I additionally use the channel as one of mutual communication, answering issues and questions operating departments have or sharing learning resources.
What projects have stood out to you the most?
Creating the Anti-Financial Crime Department’s network analysis tool. It makes suspicious behavior in our P2P (peer-to-peer remittance) feature clear at a glance, enabling a prompt response.
Refund scams abusing P2P have been on the rise, and this tool lets us identify user accounts likely to be involved in such criminal activities with a single look. Combined with follow-up investigations, this makes it possible to find evidence of money laundering. It is an extremely effective tool in creating safety and peace of mind for those using PayPay, and made the Anti-Financial Crime Department very happy, also leading to the filing of a patent.
It seems PayPay’s data and AI usage is moving forward in full sprint. What issues await you on this path?
There are two issues, broadly.
The first is not being able to keep up with the plethora of areas that could use data analysis or machine learning. On the flipside, that means plenty of room for achievements. We’re in the process of spreading machine learning and data analysis internally through introducing them across many teams, making explaining and proposals very important. We have to take a proposal-based stance, giving advice on where to start in response to even hazy questions about whether AI can solve certain problems.
The other issue is improving the accuracy of machine learning models. Our predictions often miss the mark, even when we expect adding a new feature to add a certain level of accuracy. We do need a certain degree of expertise and experience, but trial and error through experiments is unavoidable. That’s what is both difficult and fun about this field.
Growing to Reach New Heights Through Teamwork
What is your dev environment like?
I use a MacBook Pro as my local machine, but hardly ever its CPU power. Most of my work is done on Vertex AI on Google Cloud Platform. A fully GCP environment is very efficient, as I often use BigQuery data.
How do you make progress in your work?
Initially, I often gathered info on issues, proposed analyses, analyzed, and gave reports alone, but hope to create more forums to get opinions and checks from those around me with the recent increase in team members.
When analyzing alone, you tend to become self-satisfied, and it’s difficult to notice your own mistakes. Since we have a treasure trove of data analysis pros with Kaggle experience, I hope to expand our horizons through propping each other up to increase accuracy and implement new ideas.
What are the highlights and rewards of working at PayPay?
Looking across the company, there are still a lot of areas yet to implement data analysis and ML. As such, the highlight to me is being able to help people wherever I go. Also, PayPay is active in filing patents, and several have been filed for projects I’ve led. It’s very rewarding to do work that leaves a legacy.
The Endless Potential That Awaits Beyond Trial and Error
What do you want to achieve moving forward?
I plan to promote the utilization of AI broadly throughout PayPay, but my vision is of a service that addresses pain points ahead of the curve. I want to deliver services that predict preferences and life stages based on everyday behavior and meet user needs, and that enable preventative measures through predicting crimes before they happen.
I would love to create a service nobody has ever conceived through natural language processing like LLM, while also refining analyses by using group companies’ data. Being in the dawn of the AI era gives us infinite possibilities, and I plan to plunge into them.
What do you value as a data scientist at PayPay?
I do like all of PayPay’s 5 Senses, but as a data scientist I especially value “Be sincere to be professional,” which preaches being an uncompromising and sincere professional, creating new opportunities and value, and following through.
Grappling with data requires sincerity. Sometimes the results we get betray our expectations, but we can’t interpret them to suit our convenience; we have to accept them as they are. Then, only by digging for the reason behind the results, can we arrive at the truth. You need an unbending will to continue trial and error for as long as it takes.
And to top it off, a message to our readers, please!
PayPay is pioneering a new world of finance through new, never-before-seen initiatives. Anyone who can enjoy the pains of forging a new path will excel here. I look forward to meeting anyone willing to discourse not only with data and papers, but also people, AI, and even truth itself!
Current job openings
*Job openings and employee affiliations are current as of the time of the interview.