I architect integration layers and the AI-native pipelines that build them.
From spec to pull request. I build the workflow, then let it run.
See the live pipeline metricsAbout
I'm a software architect with 25+ years across data, infrastructure, and system integration. Two things define my work today: architecting the integration layer that ties a multi-application e-commerce platform together, and designing and operating an AI-native delivery pipeline that takes work from spec to pull request with public metrics.
At Davidson's I architected and maintain the REST integration layer connecting the e-commerce system's applications: a unified API over the ERP, dealer data, product catalog, and storefront systems, with cross-system orchestration where dealer workflows need it. I also manage two offshore development teams and own daily scrum. When a teammate's weekly pricing audit was eating hours, I automated the entire thing: email parsing, Playwright-driven login and price comparison, and distribution of the results. That's the kind of problem I look for.
My development approach is AI-native. I built Cornelius, a custom agent framework on Claude Code that dispatches Claude Code, Codex CLI, and OpenCode workers behind one orchestration contract, with per-task cost tracking against a model pricing table. It runs the Davidson's delivery workflow and it built this site.
I'm based in Graham, NC. I do my best work with a hard problem, clear ownership, and the autonomy to find the path.
Governed by default
Every agent tool call is permission-gated. Secrets are injected at runtime from Bitwarden Secrets Manager and never written to disk. Every dispatch writes to a trace ledger, and a human merges every pull request. Retries are bounded and classified: a verification failure re-dispatches the phase, an infrastructure fault marks it blocked and hands it back to me, because a retry will not heal a missing tool.
Projects
-
Cornelius, AI-native delivery framework — 2025 to present
Spec to plan to PR pipeline across ten projects, multi-runtime dispatch, governance built in.
-
Industrial data pipeline — Pureflow, 2023 to 2024
Python system-layer integration ingesting data from remote PLC systems into MySQL, 40+ million data points, with fully automated quarterly customer reporting.
-
Home automation and data pipelines — 2025 to present
Weather station collector with idempotent upserts and MQTT discovery into Home Assistant; daily irrigation recommendation from reference evapotranspiration and drought-monitor data, gated by live soil-moisture sensors. Shipped through the same Cornelius pipeline used at work.