At JAPAN AI, the Infrastructure Team plays a crucial role behind the scenes, ensuring that every aspect of our platform runs securely, reliably, and at scale. Their work ranges from cloud resource provisioning and CI/CD tooling to network optimization, observability, and the development of critical platform features like authentication and billing.

In this interview, we talk with Joshua Martin, the Infrastructure Team Lead at JAPAN AI. With nearly a decade of experience working across the UK and Japan—including several years in AI-focused startups—Joshua shares his personal career journey, the mission of his team, the unique challenges they tackle, and what he finds most exciting about building the infrastructure for modern AI products.

Please provide a brief self-introduction and your background.

I've been working as a software engineer for about ten years. I graduated from university in England in 2015, majoring in video game programming rather than computer science, though around 75% of my classes were shared with the CS track. This meant I still got a solid grounding in fundamental software engineering.

After graduating, I joined an online survey company on the south coast of England, where I focused on backend development using PHP. I spent about four years there before deciding to take a big leap—I applied for an infrastructure engineering role in Tokyo. It was an unusual situation because the company that hired me actually closed down right as I arrived, so I didn’t even get to work there. But I quickly found another position with a game marketing company in Tokyo, where I worked as full-stack engineer engineer for about two years.

Following that, I joined an AI startup that built its own in-house model to analyze body photos for measurements used in healthcare and fashion. I was especially drawn to the fashion industry angle. My role was to architect and build the main product for that sector, which ended up being used by millions of people every day—a very satisfying achievement.

What aspects of JAPAN AI resonated with you and led you to join the company?

I've been a heavy user of GitHub Copilot since it first launched. Even before big-name models like GPT-3.5 and Claude exploded in popularity, Copilot was already showing how useful this technology could be for developers. I also experimented with LLama models on my local machine for personal projects.

What really stood out about JAPAN AI was how transparent the company felt. During the interviews, everyone gave me clear, consistent, and logical explanations about which industries they were targeting and why. That’s not always the case with startups; often they know what they’re building but not exactly who the customer is. With JAPAN AI, there was none of that uncertainty—they clearly understood their market and communicated that very honestly.

Also, I saw a chance to keep growing my own knowledge in what is arguably the fastest-growing field in tech today. It was appealing to think about working in an environment that uses modern, stable tooling to build cutting-edge products at speed.

Were there any particularly memorable tasks or projects you've worked on since joining the company?

Since joining, I’ve worked on a huge variety of things. The Infrastructure Team’s work isn’t always directly visible to users—it’s often about reliability, platform capabilities, and making things easier for other engineering teams.

One memorable project was implementing enterprise SAML authentication. Enterprise customers are a key target for JAPAN AI, and single sign-on with SAML is practically required for those kinds of clients. The protocol itself is well-defined, but testing it in the real world is tough. Not all providers offer usable test accounts, and even when they do, they don’t always follow the spec perfectly. We had to research which providers our customers would likely use in Japan and ensure we could test those integrations reliably. It wasn’t the most technically complex project, but the value to the business was crystal clear.

Another big project was redesigning our billing system. As AI usage was evolving so quickly, our previous usage tracking and billing calculations needed to be more flexible and ready for future changes. We had the chance to completely re-architect that system, making it both more efficient and easier for other engineering teams to integrate with. It was really satisfying to build something so foundational.

Please tell me about the development work your team is doing and your role within the team.

The mission of JAPAN AI’s Infrastructure Team is to maintain the health and stability of our products and systems. Our work spans a wide range of responsibilities: provisioning and managing cloud resources, introducing new CI/CD tooling, improving network performance, and strengthening observability across the stack. We’re also responsible for building critical platform features like authentication and billing.

As the team lead, I’m here to help drive all these goals forward with the team while also serving as a central source of knowledge on building reliable, observable systems. My role is about being deeply involved in solving tough problems and making sure the systems we build are rock solid.

What are the most technically challenging parts of your work? And how do you tackle them?

Infrastructure work runs the gamut from simple tasks to really tough problems. The hardest, in my experience, is debugging production performance issues. These don’t happen often, but when they do, there’s a lot of pressure because the stakes are high.