I build software people use,
with AI built in.
I'm Jason Millis. I've been shipping apps since 2020, when I hand-wrote one in Java and published it on Google Play. Since then I've built three full-stack products, including Spatchr, a field-service CRM with an AI assistant I engineered that gets 96% of real-world requests right.
Open to junior and AI-native engineering roles · Remote or Cincinnati · Available now
Spatchr: run a service business from your phone
Leads, estimates, scheduling, crews, invoices, card payments, e-signed contracts and an AI assistant, in one app that works with no signal. Live at spatchr.com.
An AI assistant built to touch real data safely
Spatch AI answers questions like "who owes me the most?" from the business's own records, and can book jobs or record payments by voice. I engineered it to be measured and safe, not just impressive in a demo.
- Tool calling: 6 tools generated from the app's data model; 5 read, 1 acts
- Guardrails: preview → confirm → commit, role checks, a 12-step cap
- Evals: 79 of 82 real-world requests correct; every miss logged
- Cost control: about $0.56 per user per month, every call metered
- Voice: Whisper speech-to-text, hands-free follow-ups
- Security: keys stay on the server; only vetted models in production
Other things I've built
Shoptigo: one codebase, many online stores
A multi-tenant ordering platform on Vendure for local delivery and wholesale businesses. Each shop gets its own domain, branding, delivery and payments, with no deploy to add one, and attack probes in CI keep shops isolated.
G-Shop: e-commerce store for a client
A React storefront and admin dashboard with an Express and Prisma backend I wrote: a 29-model schema, stock-safe checkout in a database transaction, Stripe, shipping labels and AI shopping assistants.
Clear Exterior: business site and lead pipeline
A fast static site with a quote form handled by a Cloudflare Worker that feeds Google Sheets, plus structured data on every page for local search.
Dartscape: my first app
A native Android darts app I wrote by hand in Java before AI coding tools existed: scoring for many games, computer opponents, custom dartboard views and stats in SQLite. About 14,400 lines, published on Google Play. Rebuilt in 2026 with Expo and TypeScript.
Direct. Verify. Ship.
Direct
I make the calls AI can't: offline-first architecture, a shared business-logic core, the data model, the UI, and what to build next.
Verify
I build the checks that prove AI-written code works: 35% of Spatchr is tests, 140 architecture rules, fuzzers, and evals for the AI itself.
Ship
I deploy and run it all: Linux servers, Docker, self-hosted Postgres with backups, Cloudflare, Stripe and mobile releases.
What I work with
| AI | LLM tool calling, agent loops, evaluation suites, guardrails, cost metering, speech-to-text (Whisper), OpenAI / OpenRouter, Claude Code |
|---|---|
| Languages | TypeScript, JavaScript, Java (Android), SQL, HTML/CSS |
| Frontend | React, React Native / Expo, React Native Web, Remix, Vite, Tailwind, Redux Toolkit, TanStack Query |
| Backend | Node, Express, NestJS, REST, GraphQL, auth (JWT, Argon2id, Better Auth) |
| Data | PostgreSQL (row-level security, migrations, backups), Prisma, Drizzle, TypeORM, SQLite |
| Infrastructure | Linux VPS, Docker, Coolify, Cloudflare (Workers, R2, Pages), GitHub Actions, Expo EAS |
| Integrations | Stripe (Payments, Connect, Terminal), EasyPost, Google Maps |
| Testing | Unit and integration tests, architecture tests, fuzzing, Jest, Vitest, Playwright |
Let's talk
I'm looking for a junior developer, AI-native engineering, AI automation or AI code-evaluation role. Remote, contract, part-time or hybrid in Cincinnati. I can start immediately.