I’ve been a professional software engineer since 2005, mostly in C# and .NET. Today I’m a Staff Software Engineer at Vonage, working on contact centre technology, where I write code and lead technical work across teams.
Outside work I’m a radio amateur, callsign M0LTE, and I write open source software for amateur radio. Over the last six months, AI coding agents have changed how much of that one person can do.
At work
- Backend services and platforms. Established application platforms and cloud-hosted microservices: keeping them running, and modernising them without losing sight of where the platform is heading.
- Reliability and observability. Diagnosing complex behaviour in production, and improving structured logging and diagnostics so the next investigation is easier.
- Developer tooling. Shared tools and automated checks that catch problems earlier, such as static analysis, database schema validation and API documentation. If a problem keeps coming back, I’d rather solve it once for everyone.
- AI-assisted development. Guidance on what coding assistants and agents are good at, where they go wrong, and how to use them responsibly.
Technical leadership, for me, is mostly about clear communication, supporting colleagues, removing obstacles, and the problems that don’t fit neatly inside one team or service.
Building with AI
Twenty years of writing software is what makes the agents useful to me. I decide what to build and how it fits together, I review what they produce, and I insist on tests that hold it to the published specs, to real radio hardware and to other people’s implementations. What the agents change is the pace: projects that would have taken me years of evenings, or that I’d never have started, now take months.
Some key examples of this:
- ax25spec and ax25sdl. The AX.25 packet radio specification as editable source, with its state machines transcribed once and generated into seven programming languages. Transcribing every diagram faithfully and keeping seven code generators in step is careful, tedious work that I’d never have fitted around a day job.
- packet.net and its libraries. A packet radio node and AX.25 stack in .NET, tested continuously against the established implementations, with its building blocks published as reusable packages. The authors of the node software I test against built theirs largely alone too, over decades; this has taken months.
- APRS libraries. Libraries in C#, Rust, Python, TypeScript and C that encode and decode every APRS data type and agree with each other byte for byte. Writing one by hand is a difficult hobby project, which is unlikely to be a complete implementation; writing five, each idiomatic in its language, and keeping them identical is where the agents made the difference.
- MS110D and STANAG 5066. An HF radio modem built from MIL-STD-188-110D, in .NET and C++, with a STANAG 5066 data link on top for IP traffic and mail over HF. These are long, dense military standards full of signal processing a long way from my day job and far beyond my DSP knowledge, and getting them right takes many rounds of implementing, simulating and measuring. With agents, each round is quick enough that finishing became realistic.
The rest are on my projects page.