Skip to content
~/peter-filias/tixwork

TixWork · Founder & Agentic Engineering Lead · Apr 2026 – Present

I architected and shipped a multi-tenant platform in six months.

One engine, five verticals, a 100+ location production deployment, and an engineering workflow where AI agents do the volume and I set the direction.

// summary

Since April 2026 I've led TixWork end to end: product, architecture, delivery, and go-to-market. I designed a next-generation maintenance and work-order platform where one codebase serves restaurant, hotel, daycare, fitness, and fleet operators through tenant templates instead of forks. At the same time I kept a production system running for a 100+ location restaurant franchisee.

I ran it the way I run engineering organizations. Decisions are written down, quality gates live in CI, and work flows through a queue that both people and AI agents can execute against, with clear guardrails on what an agent may and may not touch.

// by the numbers

1,710

Commits since Apr 1, 2026 across the platform, the production app, and the marketing site.

482

Merged pull requests across 3 repos, each passing through CI gates before merge.

617

Tracked issues closed, planned and triaged in an agent-readable work queue.

5 → 1

Five verticals* (restaurant, hotel, daycare, fitness, fleet) served by one engine and one schema.

and growing as needed

// what i built

The platform: one engine, many tenants

  • Made the core architectural call: verticals are tenant templates, not forks. A new tenant is configuration (template, seed data, branding, hostname) on the same app image, so every fix and feature ships to every vertical at once.
  • One schema, one ticket model, and one asset register with stable asset IDs, cascading department → equipment pickers, and role-based home screens.
  • TypeScript pnpm monorepo: Hono API, Vite + React web app, Drizzle ORM on Postgres. Tested with Vitest and Playwright.
  • Approved design spec, implementation plan, and architecture decision records (ADRs) before scaling the build.

Production: a live system for a large franchise operator

  • Run the production PHP/MySQL application on Docker for a 100+ location restaurant franchisee, frozen to bugfixes while the new platform matures.
  • Automated staging deploys on every push, followed by post-deploy Playwright smoke tests against the staging environment.

Delivery and quality gates

  • CI spins up Postgres, seeds data, and runs the full suite plus Postgres-backed end-to-end and smoke tests on every change.
  • Demo deploys are gated behind explicit preflight checks and post-deploy smoke tests, with no write path to production.
  • The marketing site, tixwork.com, is a static site plus a Cloudflare Worker that auto-deploys on push.

// how i build with ai agents

TixWork has no AI features in the product. AI is how the product gets built. I treat agents as a team I lead: they get a backlog, a definition of done, and rules of engagement.

Agent-readable work queue

Every repo kept a structured issue log that agents parse into ready, blocked, and in-progress views, so they pick up work without a human handoff.

Triage for cloud agents

Every open issue sorted into agent-ready, agent-assisted, or human-only, and each repo audited for agent fit: CI gates, stubbed billing and email, no secrets required.

Written guardrails

An AGENTS.md defines the rules: agent PRs never target main, demo deploys need my sign-off, and agents never touch secrets or the production customer.

Facts-only content pipelines

Assistants draft growth plans, comparison pages, outbound kits, and call scripts under a strict no-invented-stats rule, with open questions flagged for my review.

// go-to-market and business

// what this says about how i lead

Stack: TypeScript, Hono, React, Vite, Drizzle, Postgres, Vitest, Playwright, PHP, MySQL, Docker, GitHub Actions, Cloudflare Workers.

// contact

Get in touch: [email protected], or see tixwork.com.