A selection of situations from sixteen years of engineering work.

Zeals was a Japanese chat-commerce startup entering a period of rapid growth. The technology stack was a Ruby on Rails monolith that could not support the planned US market launch. The team structure compounded the problem: frontend, backend, and DevOps were organized as separate functional groups rather than cross-functional product teams, which meant that any feature touching more than one layer took weeks of coordination to ship.

I joined as the fourth frontend engineer and was promoted to Head of Frontend within months. I presented a new team structure to the engineering department and secured buy-in from the CTO and VP of Product. Over the following quarter I hired ten engineers and, together with two other engineers, designed a new platform architecture: microservices, micro-frontends, and an auto-generated API gateway. A three-person team produced the full platform design and migration strategy in two months.

go2cloud was a German company that had spent most of its operational history running crypto-mining data centers. In late 2021 it acquired a French VDI startup, OmnyCloud, and found itself with a product and no established software function to develop it further. There was no software engineering team, no development process, and no software-product culture. I was the first software engineering hire after the acquisition.

In the first month I worked through the acquired codebase, onboarded myself to the technology, and integrated the VDI product into go2cloud's existing data center infrastructure. Over the following months I hired a team of seven across frontend, backend, DevOps, and QA. I redesigned the network architecture for lower-latency access across European regions, shipped a Cloud Gaming prototype on top of the VDI stack, and replaced a third-party remote-desktop streaming vendor with a WebRTC-based protocol designed in-house, which removed a third-party dependency that had been putting pressure on margins since the acquisition.

YoloPrice was building an AI-driven commerce product with a specific security constraint. The core algorithms were patented and had to be protected against reverse engineering, but the end-user experience had to remain smooth across iOS and Android, and the SDK had to be small enough that B2B integrators would actually use it.

I wrote the Android MVP myself in the first month, largely to prove that the product hypothesis was feasible at all. Once that was confirmed, I designed the production SDK as a Kotlin Multiplatform project with an embedded gRPC server running inside the mobile application itself. I implemented on-device encryption, binary obfuscation for sensitive paths, and a custom WebView for marketplace authentication that kept every piece of user data inside the application boundary. I hired two mobile engineers and worked with a DevOps engineer to resolve a release pipeline bottleneck.

Echo Analytics was a geospatial SaaS platform. Its architecture had grown into a monolith that could no longer support the product roadmap, and the original plan was to replace it with a full-scale microservices platform. Midway through the project, the company cut the engineering budget substantially. The original design was no longer viable.

Rather than scaling the original plan down proportionally, I proposed a different architecture entirely. Small, heavily automated product-services that a team of one or two engineers could take from concept to production in two to four weeks. I redesigned a core database module handling over one hundred million records, introducing better partitioning, more aggressive indexing, and a Redis caching layer for hot paths. I worked through a DevOps bottleneck and rebuilt the delivery process around the new budget and team constraints.

OneFactor was a BigData and ML services company working primarily with telco and financial clients. At the time, the company had not yet adopted Kubernetes, and was looking for a first team to pilot the adoption of Kubernetes across the organization. The new product — SmartIndex, an analytics platform for out-of-home advertising — was chosen as the pilot for that transition.

I built the team from zero, delivered the MVP in two months, and designed the Kubernetes deployment architecture with separated development, UAT, and production environments and blue-green deployments for production releases. I wrote the CI/CD templates that later became the organizational standard. The service architecture I designed — stateless, horizontally scalable, with an efficient resource profile — handled one thousand requests per second on two small pods.

Sberbank's SME B2B Banking department had an internal product that generated roughly thirty percent of the department's revenue, and a frontend team that had not shipped a release in several months because a framework migration had stalled. The product was business-critical and could not stay frozen.

I joined, unblocked the stuck release in the first weeks, and was given ownership of both the frontend and backend teams — ten engineers and two leads between them. I introduced Scrum as an actual rhythm rather than ceremony. I led a clean Java 7 to Java 17 migration on the backend, and began a monolith-to-microservices transition.