Senior Java Engineer

Java 17 and 21 LTS by default. We avoid Java 8 unless your platform forces it (and we'll quantify the cost of staying there during scoping). Frameworks: Spring Boot 3.x is most common; we also handle Quarkus, Micronaut, and (rarely) plain Servlet stacks for legacy integrations.

Yes — strangler-fig is the default pattern. We carve out vertical slices (a feature or bounded context), rewrite them as Spring Boot services, and route traffic incrementally via a thin gateway. The legacy app keeps running until we've migrated enough that retiring it is a small step.

First step is always profiling, never guessing. We use Java Flight Recorder + Mission Control or async-profiler to collect heap and allocation traces. From there, choose GC (G1 for most workloads, ZGC if pause-time-sensitive), tune heap sizing relative to container limits, and address allocation hotspots in code. We document every JVM flag we change with the data that justified it.

Not by default. Microservices solve organisational and scaling problems that small teams don't have. We do an honest assessment in the first sprint — if your monolith is fine, we'll say so and focus on modular boundaries inside it. If it genuinely needs splitting, we extract by team boundary, not by 'this should be a service' instinct.

We treat Spring Boot major upgrades (2 → 3, etc.) as their own engagement. The path: dependency analysis, security migration (often the biggest blocker), Jakarta EE namespace move if applicable, integration test pass, staged rollout. Most enterprise upgrades take 4–8 weeks of focused effort.

Real databases via Testcontainers, almost always. Mocked DB tests catch a fraction of real bugs and create false confidence around migrations. The CI cost (a few extra seconds per test) is dwarfed by the production incidents we've prevented.