Next, we spoke with Mr. Hayato Kuroda, who was recognized as a PostgreSQL contributor in January 2025. He is a world-class developer whose name is already listed on the official PostgreSQL web page — an impressive achievement for someone his age. He states his mission is "to increase the value of database OSS itself through my own development work, while also elevating Fujitsu's presence."

——Please tell us what you do day-to-day as a contributor.

Kuroda　In the PostgreSQL community, all communication takes place on mailing lists. As developers, we constantly monitor these emails, looking for new discussions or replies. If a message calls for a follow-up, we join the discussion to deepen it. When proposing a new feature, I first summarize its background and intent and then post it in English to the "pgsql-hackers" mailing list. Once various comments start coming in, we engage in interactive discussions to refine and shape the feature. Additionally, providing technical reviews of proposals from other developers is also an important part of our activities.

——You are currently focusing on developing an extension for PostgreSQL's logical replication. Could you tell us what kind of feature this is?

Kuroda　In current standard PostgreSQL replication, the standby side cannot modify data. Logical replication aims to enable this. By abstracting information and replacing it with table-level changes, logical replication intends to enable the subscriber side to also receive data while reflecting changes.

——You also gave a talk at an international conference. What was the topic?

Kuroda　It was related to logical replication, titled “Logical Replication -- Handling of Large Transactions.” In PostgreSQL's original mechanism, changes flow to the subscriber side only after data is committed, which caused significant latency when transactions were long. We improved this in two stages. In PostgreSQL 15, the mechanism was improved so the data would be transferred before the commit and then reflected once the commit occurs. And in PostgreSQL 16, we further strengthened this by introducing dedicated workers that apply received data immediately and in parallel, significantly improving performance. What I participated in developing and presented in my talk was this parallel apply.

——Has that actually been incorporated into the official version of PostgreSQL?

Kuroda　Yes. It has already been incorporated in PostgreSQL version 16. It was the result of a team effort.

Mr. Kuroda, who is active globally as a contributor, says that obtaining OSS-DB Gold certification was a major turning point. He frankly recounts that when he began community activities in earnest in 2022, he struggled to keep up with discussions among overseas developers.

Kuroda　Honestly, I could not understand at all what was being discussed in the posts to the mailing list. Beyond the coding itself, I struggled quite a bit — not being able to keep up with high-level technical discussions with team members, not being able to deeply understand review comments from developers around the world.

While feeling my own lack of knowledge about PostgreSQL, I discovered OSS-DB Gold certification as a way to learn systematically. OSS-DB Gold certification has sample exam questions with explanations that are publicly available, and there are also study materials and books for it. I thought that learning by following this approach would be the most efficient method, and I set my sights on obtaining it.

——How did you approach your learning process?

Kuroda　It involved three steps: "solving sample questions," "reading the documentation," and "reading the source code."

First, I thoroughly worked through the sample questions and official question sets published on the web. This allowed me to objectively identify which areas are frequently tested and which fields I needed to improve my understanding of.

Next, for the parts I couldn't answer correctly, I read through PostgreSQL's official documentation. PostgreSQL's documentation is very carefully maintained, and because it is also translated into Japanese, it was easy to read through even highly specialized content.

——How did you approach the step of reading source code? The entire codebase probably exceeds a million lines.

Kuroda　At first, I was just guessing based on the file names to find what I was looking for. However, as I progressed, I realized that the directory structure is organized by component. For example, under the source there is a directory called “backend,” and within that, there are further hierarchical directories like “replication” and then “logical”. I used this structure as a clue to navigate and understand the code.

——Specifically, what kind of processing did you investigate in the source code?

Kuroda　For example, I used the source code to examine the flow from when a hot standby instance starts up until it actually begins receiving WAL records, and conversely the process of starting to send WAL records based on requests from the standby. This not only enabled me to consistently understand PostgreSQL's internal workings, from the specific flow to its external specifications, but also exposed me to its fundamental design philosophies and coding conventions. This was a major learning experience that directly contributed to my current development activities in the community.

——What changes have you experienced since obtaining OSS-DB Gold certification?

Kuroda　Technical concepts and fundamental designs that had previously seemed difficult became clear, and I gained confidence in participating in community discussions. As a result, I’ve been able to make higher-quality code proposals and reviews, and that has led to my current recognition as a contributor.

Also, because the number of features I was unfamiliar with decreased, even when hearing about unfamiliar technical topics in English at conferences, I am now able to think, "Oh, right, there was a feature like that," and quickly grasp the context. With experiences like these, everything tends to sink in more easily.

——For young engineers aspiring to participate in community development, what would you recommend as a first step?

Kuroda　First, read the documentation and try to find any issues or areas for improvement. This could even be something like an unclear explanation or a typo. It is very rewarding when your contributions are accepted by the community, and that success can be a great motivator.

——What are your goals for the future?

Kuroda　I hope to establish a position where Fujitsu’s leadership is widely acknowledged in the field of logical replication that I am currently working on. Furthermore, I aim to be recognized as a major contributor or committer. Because I believe that consistency is the most crucial factor for gaining trust among developers in the community, I will continue to engage in high-quality development.

Also, by leveraging the systematic knowledge I gained through OSS-DB Gold certification, I would like to expand my scope beyond specific functional areas to cover a wider range of domains.

The third interviewee is Mr. Koshi Shibagaki, who is in charge of feature development for Fujitsu Enterprise Postgres. He joined Fujitsu in 2022 and is 28 years old. He is a young engineer who obtained OSS-DB Silver certification in his first year and Gold certification in his second year.

——Please tell us about your current work.

Shibagaki　I am in charge of development of Fujitsu Enterprise Postgres. While leveraging PostgreSQL's strengths as an open-source solution, I implement functional modifications and performance improvements to its core, and develop proprietary extension features to meet our customers' diverse needs. In recent years, generative AI — and methods like RAG in particular — has attracted attention, and the need to leverage business data with generative AI has been growing. Accordingly, databases have evolved from being mere storage locations to playing a role as a knowledge base for AI to reference. In light of these changes, we are focusing on developing features specifically aimed for the use with generative AI.

——What specific features are those?

Shibagaki　To safely leverage generative AI in business, we are incorporating features that, while minimizing the burden of setup and operation, allow data stored in the database to be connected directly to generative AI usage.

In RAG, internal company-specific data is converted into vector data, relevant information is extracted through similarity search, and those results are passed to the generative AI. The conversion to vector data requires processing such as passing data to an embedding model using Python libraries and performing tokenization, which can be somewhat complex. If these were implemented as tools on the application side, the burden of development and maintenance would increase. So, we provide this vectorization processing as a database feature. Inside the database, dedicated worker processes operate, and they acquire data at the time it is stored or modified, and automatically execute everything from feeding it into the embedding model to vectorization and storage into tables. Also, texts entered during similarity searches are vectorized by the same mechanism. The application side can use semantic similarity search simply by specifying text, without being aware of the vectorization.

As a result, vectorization is performed automatically at the time of data storage, and because it is automatically updated when data or query text changes, the labor required for preprocessing, such as ETL, can be significantly reduced. A major feature is the ability to consolidate preprocessing for leveraging generative AI in business applications on the database side. Also, the data handled by this feature is protected by existing security functions of Fujitsu Enterprise Postgres, including Transparent Data Encryption, allowing it to be used as safely as conventional business data.

*Vector data: Data that represents unstructured data such as text or images as arrays of numerical values (vectors) in hundreds to thousands of dimensions. This enables mathematical calculation of semantic proximity between data (similarity search).

*Embedding model: A machine learning model for converting data such as text into vector data. It captures the semantic features of the input text, converting semantically similar sentences into close vectors, and semantically distant sentences into distant vectors.

*ETL: Abbreviation for "Extract, Transform, Load," which refers to a series of data processing steps where data is extracted from its original source, transformed into a format suited to the purpose, and loaded into the target database or data warehouse.

——It’s quite fast to obtain Gold certification in your second year with the company.

Shibagaki　Originally, I thought it would be good if I could get it by the end of my third year. However, seeing my peer obtain it in their first year changed my perspective.

I began to think, "I should also acquire a deep PostgreSQL knowledge and solidify my foundation for development much sooner."

——Please tell us about your study method.

Shibagaki　Rather than simply memorizing commands and configuration items, I set my goal as being able to explain in my own words "what kind of processing is running behind that command" and "which process is responsible for what role."

For command-related areas, since I had many opportunities to execute them in my work, it didn't take too long to improve my accuracy. On the other hand, my understanding of the internal structure behind command execution was initially weak, and that was my challenge. Tuning and troubleshooting are areas where I had few chances to encounter them in my work, and they require a lot of knowledge and experience. So, this is where I particularly struggled.

It happened to be just when the official study materials had come out, so I first went through those and then deeply explored the parts I didn't understand. After grasping the background and overall picture with the official documentation, I organized my knowledge with books and LPI-Japan-endorsed materials, and finally checked "why this behavior occurs" in the source code — in that order, I deepened my understanding.

——Why is it necessary to delve into the source code for certification exam preparation, as it seems like a high hurdle?

Shibagaki　If passing the exam is the only goal, study materials alone might be enough. However, we are not just users; we are developers. After passing, we have to create things. In that case, it is better to deeply understand how things work behind the scenes. That is why I took a study method that went as far as reading the source code.

——What changed after obtaining OSS-DB Gold?

Shibagaki　The biggest change is that my PostgreSQL knowledge connected not as dots but as a line. With a systematic understanding, I became more actively involved in technical discussions in development work and could offer more constructive opinions. For example, when discussing the behavior of a specific function, I could offer my opinions while deeply considering the performance impact and coordination with other functions, based on my understanding of the internal structure.

Also, even when I face unclear points or challenges, I can now accurately pinpoint, "For this function, there might be a hint in this section of the official manual, or around this part of the source code." This has significantly reduced the time taken to solve problems, and also has led to improved development efficiency.

——Please tell us about your future career.

Shibagaki　In addition to product development, I also would like to contribute to the OSS community. Improving our own products and contributing to the community are by no means conflicting goals — rather, I believe they are closely intertwined.

If upstream issues are resolved in the community, it directly leads to improvements in product quality. I want to give back the insights gained in product development to the community and make concrete contributions such as bug fixes, documentation improvements, and sharing of verification results. I hope to experience the satisfaction of knowing that features I have created are being used worldwide.

Finally, we asked Mr. Awatsu for a message directed at executives of other IT companies.

Awatsu　Customers will increasingly seek open, de facto standards and solutions. Even among large-scale customers, multi-vendor environments are increasing, and a trend is emerging where they are establishing standard specifications rather than adopting products from each vendor individually. It is OSS that is chosen as that standard. I believe that leveraging OSS should be the first consideration.

Mr. Awatsu also spoke as follows to young engineers:

Awatsu　If you only focus on vendor-specific certifications, your progress will be limited. Open technologies are a common denominator and will be in even greater demand in the future. The need among customers to avoid vendor lock-in will only grow stronger.

We also asked Mr. Kuroda and Mr. Shibagaki for messages to those who aim to obtain OSS-DB Gold certification.

Kuroda　I believe OSS-DB Gold is both the best textbook for PostgreSQL and the shortest path to catching up with developers worldwide. Through studying for the certification exam, let’s accumulate the knowledge needed to discuss with engineers at the cutting edge of modern technology.

Shibagaki　I think OSS-DB Gold is a certification exam that delves into profound content, where superficial knowledge or rote memorization simply won’t suffice. That is precisely why the sense of accomplishment when you obtain it, and the depth of the knowledge you gain, are immeasurable. It is important to take it on with deep understanding of "why things are the way they are" and "how they work internally." Reading the official documentation, systematically organizing knowledge with books, and finally confirming the operating principles in the source code —— I believe this multi-faceted approach will reliably lead you to success.

Kuroda　PostgreSQL's support period is also five years, so by the time five years have passed, the knowledge may have become outdated. I think it is correct to say that one needs to learn again. For those actually developing, we are constantly tracking the latest information daily, so it becomes a good opportunity to review a broad range of knowledge.

What was particularly impressive in this interview was the fact that even world-class engineers, like PostgreSQL contributor Mr. Kuroda, found OSS-DB Gold indispensable. Mr. Kuroda stated, "It enabled me to propose and review higher-quality code, leading to my contributor recognition," and Mr. Shibagaki expressed his feeling that "knowledge connected not as dots but as a line, greatly shortening the time to solve problems."

It is precisely the most exceptional individuals who seek to understand the whole picture. And for that pursuit, OSS-DB Gold stands as the paramount pathway.

Company Overview

■Company name：Fujitsu Limited
■Founded：June 20, 1935　
■Headquarters：4-1-1 Kamikodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa 211-8588, Japan　
■Business description　
・Service solutions
・Hardware solutions
・Ubiquitous solutions
■URL：https://global.fujitsu/en-global