The proof, not the pitch
Pilot for Trades — a full field-service platform for electrical contractors: quoting, dispatch, invoicing, payments, QuickBooks, a voice-AI assistant, and mobile. Built to pressure-test how far AI now lets one senior engineer go. This page is the receipts.
What it is: real, production-grade software — a web app, a mobile app, and infrastructure-as-code, with live integrations (Stripe, QuickBooks, Telnyx, Twilio, AWS Cognito) and a local, $0-inference voice-AI stack (Whisper + Llama 3.1 8B on one GPU). Every figure below is measured from the repository, the test runners, and git history.
What it isn’t: a business. It’s an R&D build — never released, no customers, by design. It was architected to the SOC 2 control set — built to the controls, not audited. Stripe is built and processing charges; it has not been launch-verified at scale. Coverage is ~71–76% measured — not 90%+.
The point was never the product. It was proving the method I bring to your team: one senior operator directing twelve specialized AI roles across six models — and owning the judgment calls that are expensive to get wrong.
Measured from the repo, the test runners, and git history — solo, AI-augmented, in about six months.
252K
Lines of application code
189
PostgreSQL tables
952
Design & architecture docs
~8,000
Automated tests · 0 failing on a clean run
71–76%
Measured test coverage · full-stack
77
API modules · ~720 operations
782
Commits · 156 active days
10+
Wired integrations
100%
Solo · ~6 months
Not by typing faster — by running it like an engineering org where the team is AI. Twelve specialized agent roles — architect, backend, frontend, QA, DevOps, integrations, voice-AI, UI, domain expert — each with its own rules and remit, backed by six different models routed by what each is best at.
The headline isn’t that AI wrote the code. It’s that one experienced person can now direct a specialized AI team — set the architecture, assign the work, review the output, keep a dozen workstreams coherent — while owning every call that’s expensive to reverse. That’s the same way I work inside a client’s org.
Take the AI-First Readiness Scorecard — 18 questions, about four minutes. See your readiness band, your six dimension scores, and what to fix first. Free.