I’m Tony Erwin, a software architect at IBM.

I build the platforms other engineers build on, in environments where getting it wrong is expensive. These days that means agentic AI — multi-agent systems, the protocols and libraries that let teams plug into them, and the governance that makes any of it safe to ship. Before that it was regulated cloud infrastructure and the IBM Cloud console, which I helped grow from its first lines of code into a distributed system serving clients worldwide.

The through-line is the same either way: reusable architecture, built to be depended on, in places where reliability and trust are not optional.

I write here about the parts that don’t make it into the documentation — the architectural decisions, the tradeoffs, and the things I’d do differently. The hard problems in this space are ones the whole field is still working out, and writing about them is how I think them through.

You can find me on LinkedIn and GitHub.

Views expressed here are my own.