Technical writing on AI systems and the infrastructure under them, payments engineering, release management, and quality automation.
Release and QA timelines cause branches and instances to drift, turning one product into multiple codebases and forcing duplicate fixes. Code duplication, asset synchronization, configuration drift, database compatibility, duplicated testing, and release management as a stream of its own. All of it traces to one root cause, and multi-tenancy removes it.
Read the article →Fifteen decisions to settle before the first endpoint ships, each stated as the problem it solves. RSA-signed JWTs so any service can validate without being able to forge, scopes enforced centrally, tenant isolation in four layers with a compile-time guard, cursor pagination, idempotency keys that fail closed, and errors that separate the caller contract from the on-call signal.
Read the article →The format is only half the story. The interesting part is the asynchronous pattern the standard assumes, and everything an acquirer has to build to live with it: correlation keys, timeouts and reversals, persistent connections, and an async request-reply correlator that still works once you run more than one instance. Most of the design is not about the happy path.
Read the article →Having worked with all of them, there is no architecture that is always the right answer. Start with a monolith with strict module boundaries, keep the domain separation without the network calls, and extract a service when there is an actual reason. Service extraction is not a failure of the monolith, it is an evolution of it.
Read the article →Complex problems rarely fit inside one technology. Tell us what you are trying to solve and we will tell you how we would approach it.
Discuss a Technology Challenge