Updated 27 August 2026 - Dallas, Texas
I’m a Senior Technical Program Manager at Microsoft, working on digital platforms and data infrastructure. That’s the day job and it’s a good one.
What I keep thinking about, though, is smaller and stranger: much of the AI tooling conversation is happening around Python and JavaScript, while a huge amount of the software running banks, retailers, hospitals, insurers and other large enterprises is still Java. The Java ecosystem feels underrepresented in this new wave, so I’ve been spending my time exploring that gap.
Right now, that means three things.
I wrote a GitHub Copilot skill for Helidon, and it was merged into GitHub’s Awesome Copilot project. I’ve also written twice about this work for Oracle’s Helidon publication.
I’m building a multi-agent demo in Java using Helidon 4, LangChain4j, and models running locally. One idea I want to test is whether Java’s virtual threads are particularly well suited to agent workloads, where so much of the execution time is spent waiting on I/O. I’m presenting the work at JavaMUG in Dallas on 9 September, which gives me a useful deadline.
I’m also a PhD student in AI at Capitol Technology University, looking at how these systems can be managed and scaled responsibly.
What I’m saying no to
Nearly everything else.
New frameworks, new side projects, clever domain names, whatever model came out this week.
I have a long history of getting interested in too many things at once, and the honest fix has been to ask one question:
Does this make the thing I’m already building better?
And to accept “no” as an answer more often.
If you want to talk about Java, agents, or what it takes to ship AI inside a regulated enterprise, I’d like that.
Most other things will have to wait.