Work with me.
An honest page about what I take on, who it suits, and how to start — written for teams who want an engineer who will own the hard part and see it through to production.
I take on work I can own end to end — and I stay with it until it runs dependably in production, not just until the demo works. If you're building something that has to hold up, a paragraph about it is the best place to start.
Write to me →How I work
I work the way I always have: I take ownership of a build and stay with it — shaping the architecture, writing the code that carries the weight, and standing behind how it runs once real users depend on it. Thirty years of delivering systems that had to keep running sits behind that, much of it in relationships measured in years, not weeks. Most engagements take one of these shapes.
- 01
Own the build, end to end
A senior engineer who takes the system from requirements through code, deployment, and the production hardening most people stop short of — retrieval pipelines, agent orchestration, evaluation, observability, all of it built to hold. I don't hand the work off at the architecture; I write it, and I stay with it until it runs.
- 02
Lead a modernization
Taking a legacy estate to a modern foundation without stopping the business — lift-and-shift where that's right, real re-architecture where it pays off, and a steady hand on the cutover. I've moved federal and enterprise systems this way for years, historical data and all, and stayed on to keep them running.
- 03
Architecture, with the delivery behind it
A clear, honest read on a system you're about to build, scale, or replace — and, when you want them, the hands to build what the read recommends. Not a memo that ends at the door: the thinking and the implementation come from the same person.
- 04
Pre-flight review
A grounded check before an AI system meets real users — the failure modes, the evaluation, the cost, and the operational details that quietly decide whether a launch holds. Where it helps, I stay on to fix what the review turns up rather than leaving you a list.
Practice in the agentic era
What, not How.
For most of my career the craft was the point. You learned the language, the framework, the runtime, and you built the thing yourself, line by line, because that was the work — and there was quiet pride in doing it well.
The tools changed; the discipline didn't. I design the system, set the contracts and the tests, and let AI coding agents do more of the typing. What to build and how it should hold together is mine to decide. How every line gets typed is not. The result is more software, built more carefully, with an architect's hand on every decision that matters — and the same standard underneath it all.
Daily tools — Claude Code, Codex, GitHub Copilot, MCP servers, custom skills.
I still write code most days. But the leverage is no longer in how fast I can type — it is in how clearly I can describe the work, and how rigorously I can verify the result.
Who I work best with
I'm not the right fit for every engagement, and I'll tell you so early. The ones that go well tend to share a few things.
- Teams already shipping
- I add the most where engineers are already moving, and I join as one of them — owning a hard part of the system through to production, not hovering over it with advice.
- Consequential or regulated domains
- Healthcare, finance, public-sector, transportation. The work I care about most is where the answer matters and the system has to keep running long after launch.
- Where one senior engineer changes the outcome
- Places where an experienced hand, owning the right piece of the work, genuinely moves the result — clear scope, real accountability, and outcomes you can point to.
What I will not do
- Slide-deck-only consulting. I work in code, in architecture, and in plain writing — and I stay to see it land.
- Ungoverned AI. If it moves money, grants access, or touches care, it gets deterministic code and an audit trail — not a prompt.
- Work where I can't honestly move the outcome. I'd rather point you to someone who can.
Selected work
I keep client names private, out of respect for the people the work serves — but the scale and the stakes are the part that matters here, so those I've kept. The two named entries are my own products; everything else is client work. A mix of recent AI builds and the larger deliveries behind them.
-
ClinSupplyCompass — a clinical-supply planning platform I built and operate as a product: Python, FastAPI, and PostgreSQL with row-level security, deterministic forecasting proven against the baseline workbook, audited re-planning, and AI narratives grounded in the numbers.
live · ongoing
-
Own the build
— Regulated enterprise · AWS GovCloud
Retrieval-augmented conversational AI on Amazon Bedrock Knowledge Bases, OpenSearch, and Lex, fronted by a governed LLM proxy handling authentication, model routing, fallbacks, rate and budget controls, and token-level cost — in production, with answers grounded to source and an audit trail.
recent
-
Lead a modernization
— Federal grant estate
Led the full migration of a national grant-management estate from on-premise data centers to AWS — applications, integration tiers, and years of historical data — then re-architected the legacy core onto Spring Boot and container services with secure, least-privilege access.
multi-year
-
Own the build
— Metropolitan transit & energy
Delivered transit and energy integrations that unified telemetry from thousands of vehicles and charger stations into monitoring, predictive maintenance, and work-order automation across the estate.
delivered
-
Lead a modernization
— Cross-border mainframe
Modernized COBOL and CL mainframe batch estates across two countries into optimized Spring Boot services, cutting batch runtimes from hours to minutes.
delivered
-
HeyLayla — what began as Bedrock AgentCore research became a shipped product: a private, verified matrimonial platform on a fail-closed TypeScript core, deny-all row-level security, an append-only audit log, and a Claude-powered guide that only ever assists.
live pilot
How to start
If you can describe the problem in a paragraph — what's hard about it, what you've tried, what success would look like — I'd be glad to read it.
- 01
Send that paragraph by email. No forms, no calendar links.
- 02
I usually reply within a few days, and I'll be straight with you if it's outside what I can usefully help with.
- 03
If we're a fit, we'll set up a short conversation to go deeper. If not, I'll often point you toward someone who is.