Systems rarely fail inside a single service. They fail at the seams, where services, platforms, and legacy software meet, and where problems are hardest to see.
I help engineering teams find those problems, trace them to their cause, and simplify what’s creating them.
Capabilities
Services can function exactly as designed, while the system built with them fails.
The parts of a system can behave as expected while making every future change harder. An architecture that made sense five years ago can start to get in the way as user needs change. A production problem can appear in one service while the underlying cause is in a completely different, non-obvious place.
This is exactly where my expertise is strongest.
APIs, services, data flows, event-driven systems, legacy interfaces, cloud platforms, and proprietary technology.
I bring things together so that independently evolving systems can cooperate without unnecessary exposure of their internals.
Legacy doesn't necessarily mean obsolete, and modernization doesn't necessarily mean rewrite.
I help teams find the constraints and old decisions that are getting in the way of success, and help change and evolve systems incrementally. This allows them to stay operational and useful while under development.
Distributed systems suffer from distributed issues.
I use tracing, metrics, logs, targeted testing, and architectural analysis to follow behavior across service and system boundaries, then turn ambiguous performance and reliability problems into concrete engineering problems that can be addressed.
Selected work
A passenger-facing reservation API was making people wait for years of historical data before showing the handful of upcoming reservations they actually cared about. I traced the calls, separated the two paths, and let history load asynchronously. The latency-sensitive response dropped from several seconds to consistently under 500 milliseconds.
A military simulation system could only report a single view of its simulated forces. I reworked the gateway so multiple instances could run from the same implementation with independent configuration and filtering. During a live joint exercise, three instances simultaneously provided different force views into the command-and-control systems.
A proprietary ledger and the applications consuming it shouldn't need to understand each other's internals. I built an HTTP API and OpenAPI contract around the ledger, along with native iOS and Android SDKs, giving application developers a conventional interface to a very different underlying system.
About
I'm a software architect and engineer based in Orlando, Florida. I've spent more than 27 years building and untangling distributed systems: 17 of them on defense training simulation, and the rest across travel, financial technology, and healthcare.
Castanea Systems is my independent practice. When you work with Castanea, you work directly with me.
LinkedIn