about
Software engineer. Rochester, Minnesota. Currently focused on agentic systems in healthcare.
I have been shipping software professionally for a long time. Most of that work has been web applications on top of Go and TypeScript services. The last two years have shifted toward agentic systems — tool-use loops, context engineering, eval harnesses — usually inside a larger application, usually in a regulated domain where PHI and audit trails are first-class concerns.
I also build and maintain Verve — a full-stack web framework in pure Zig. This site runs on it. The /verve route demos the live features.
I work as a tech lead at a regional health system — training engineers, advising on architecture, and shipping agentic systems under HIPAA and HITRUST constraints. Outside of that I run Sirhco Development LLC, and take a small number of consulting engagements at a time.
If you're looking for the short version of what I can help with, that lives on services. The tools I reach for are on uses.
drag me · fig. 02
sirhcohow I work
Regulated first
Compliance is an architecture problem, not a checklist you run before the audit. HIPAA and HITRUST constraints shape the first commit, because retrofitting them is where budgets die.
Eval-driven
An agent without an eval harness is a demo. I want a number that tells me whether the last change made things better or worse, and observability that lets a human intervene early.
Handoff over dependency
The work is finished when your team can extend it without me. That means documentation, infrastructure as code, and engineers who were in the room while it was built.
how I think about this stuff
An agent that works is an agent whose control loop you can draw on a whiteboard. The interesting parts are not the model calls. They are the state machine around them, the tools you chose to expose, the eval harness that tells you whether the last change made things better or worse, and the observability that lets a human intervene before a small problem becomes a big one. The model is a component. The system is the work.
what I'm working through
Learning faster than I'm forgetting. Reading Leptos source for fun. Reading rig.rs source when I have to. The agentic stack churns hard — half of last spring's libraries are noise now, half of last spring's noise is load-bearing. The skill that compounds is reading code under deadline. Not reading hot takes.
Thinking about how much of an agent's value lives in the tool schema versus the model choice. Early answer: more than I expected.
Avoiding: anything that tries to solve agent reliability with a bigger prompt. The answer is not in the prompt. The answer is in the state machine.
currently
shipping agentic workflows in healthcarePart of a small team building in-tenant agent pipelines for a regional health system. Off the keyboard: wife and kids first, the Mississippi when it thaws, and coffee that takes too long on purpose.
start a conversation
Tell me what you're building →