services
I build agentic systems and the infrastructure under them for regulated environments — and I level up the engineers who inherit them.
I am not between jobs. I work as a tech lead at a regional health system, doing this daily: training engineers, advising on architecture, and shipping agentic systems under HIPAA and HITRUST constraints. Outside engagements run through Sirhco Development LLC, a small number at a time.
Production agentic systems & enterprise RAG
buildDeterministic tool-use loops, state-machine orchestration, and retrieval that cites its sources — with eval harnesses and human-in-the-loop gates by default, because in a regulated setting the interesting failure is the one nobody caught.
when you need it
You have a demo that works and no idea whether it will hold up against real traffic, real data, and a real auditor — or you're handling PHI and the compliance conversation is arriving faster than the architecture.
what you get
A control loop you can draw on a whiteboard, retrieval with a documented data path, an eval harness that tells you whether the last change helped, and the observability to intervene before a small problem becomes a big one.
Team enablement & agentic tooling
enablementPractical, zero-fluff training on driving Claude Code, Codex, Antigravity, and terminal agents — and building the custom agentic coding harnesses (and extending open-source ones) that make agentic tooling safe to hand a whole engineering org. Delivered internally to date; opening to outside teams.
when you need it
Your team is using AI tooling by instinct and you want them using it by method — or you want to stop buying agents and start building them.
what you get
Engineers who can build, debug, and extend this themselves. The point is that you stop needing me.
Regulated & mission-critical cloud
advisoryGCP architecture built strictly for HIPAA and HITRUST compliance from the first commit, FinOps-conscious cost control, and Go/Rust backend modernization for services where an outage or a cost surprise is not a minor incident.
when you need it
The bill stopped making sense, a service that used to be fine is now the reason deploys are scary, or the compliance conversation needs an architecture, not a slide.
what you get
In-tenant infrastructure as code, a written read on where the money and the latency actually go, and a sequenced plan your own team can execute without me.
How it starts
01
Intake
The form on the contact page, or an email. What you're building, what's in the way.
02
Scoping call
45 minutes, free. You leave with a written read on the problem whether or not we work together.
03
Written proposal
Scope, milestones, and a fixed fee or weekly rate. No hourly billing.
Engagement models
01
Production AI audit & architecture sprint
1–2 weeks. A structured pass over latency, cost, eval coverage, and security posture — you get a written read on where the system stands and what to fix first, in priority order.
02
Developer acceleration workshops
Hands-on training for senior teams adopting Claude Code, Codex, Antigravity, and custom agent-harness workflows — built from what I already run internally, not a generic curriculum.
03
Fractional AI architect / embedded lead
Ongoing, part-time architectural ownership — guiding a multi-agent system from zero to production deployment without becoming a full-time hire.
What I don't do
- —Staff augmentation — I am not a seat you rent by the month.
- —Long-term full-time contracts. The day job is the day job.
- —Greenfield mobile without a designer already involved.
- —"Make it AI" without a problem statement underneath it.
Rates depend on scope and are quoted in the proposal. The intake form asks for a budget band so the first conversation starts in the right place — "not sure yet" is a real answer.