shipping
Tome
Self-hosted ebook & audiobook platform — Node/Postgres server + Flutter client.
TypeScriptNode.jsExpressPostgreSQLSupabaseFlutter+2
nowTome · v0.12.3 — paginated EPUB reader · 2d ago
I build production things
across web, mobile, infra, and AI agents.
QA engineer by day. Force-multiplier with AI by night. 10+ years shipping. 10 projects in the registry, 4 actively under development, two of them live on the App Store and Google Play.
10+
years shipping
10
projects live
3
actively shipping
2
in app stores
· selected work
Things I've actually shipped.
Each tile is something that runs in production — on a phone, on a server, or on the wall of my house. No demos, no proofs-of-concept.
shipping
Venture Quest
Gamified scavenger-hunt app shipped to App Store and Play Store — 100+ active users.
FlutterDartFirebaseCloud FunctionsGoogle MapsFCM
shipping
Confidential Client SaaS
clientMulti-tenant marketing operations platform with AI image gen, CRM, and pipeline.
ReactTypeScriptViteHonoSupabasePostgreSQL+3
Booking time + meeting link on pipeline
active
Squadron Klaw
Open protocol for agent-to-agent communication — "SMTP for AI agents."
TypeScriptNode.jsHonoEd25519SQLiteTurborepo+1
maintained
Daily Team Mini
Multiplayer word-puzzle game with real-time WebSocket sync.
Next.jsReactTypeScriptPartyKitTailwind CSSFramer Motion
maintained
Money Allocator
Biweekly paycheck allocator that respects safety buffers and upcoming bills.
ReactTypeScriptViteTailwind CSSSupabaseRecharts
maintainedsince 2024
Homelab
Proxmox cluster + HomeKit/Matter mesh that runs the house and hosts my projects.
ProxmoxDockerHome AssistantHomeKitMatter / ThreadAqara+2
infra · automation
· the constellation
The projects are
connected.
The homelab hosts Tome. The AI image pipeline I built for client work informs how Squadron Klaw thinks about agent payloads. The multiplayer puzzle game shares Supabase DNA with the money allocator. Nothing here is a one-off — each project teaches the next one.
tip tap or hover a node to trace its connections.
Three bands — platforms & agents up top, apps & games in the middle, infra & automation at the base. Lines are real dependencies between them.
platforms
apps
infra
· track record
I build the test rig
that should exist.
Ten-plus years in QA — I've founded a QA org from zero and led engineers onshore and offshore, rising from Senior to Lead to Manager. The through-line: when the tooling to test something properly doesn't exist, I build it, often in hardware, then hand the team something that makes everyone ship faster.
10+ yearsTest automationHardware-in-the-loopTeam leadershipAI-augmented QA
· the signature move
I build robots to test things.
2019: a credit-card-swiping robot at 365 — Raspberry Pi, Arduino, 3D-printed parts — that cut manual card-reader regression by ~90%. 2022+: Raspberry Pis simulating BLE device events at Monovo so the full device-to-app flow tests itself. If it can't be tested in software, I build the hardware that can.
- MonovoQA Manager (from Senior QA)2022 — present
Founded QA from zero. Now I lead ops.
Stood up the QA function from nothing across web, mobile, and Bluetooth hardware, hired the team, and set the standards. Then I kept going — rigging Raspberry Pis to simulate BLE device events for true end-to-end coverage, and building the internal platform that became the operational backbone for multiple teams. Support resolution is ~40% faster. Nobody asked me to.
- 365 Retail MarketsQA Engineer → Senior → Test Lead2017 — 2022
Led mobile QA for the kiosks. Then I built a robot.
Led a team of QA engineers owning the nanomarket and picomarket self-service kiosks — the Android POS people use to scan and buy snacks. Manual card-reader testing was slow and error-prone, so I engineered a credit-card-swiping robot (Raspberry Pi + Arduino + 3D-printed parts) and wired it into the automation suite, cutting manual regression by ~90%. Years before AI.
see the robot → - SaltStackQA Engineer · acquired by VMware2015 — 2017
Open-source QA at startup speed.
Automated regression with Selenium and triaged community bug reports straight from GitHub for a fast-moving infrastructure startup — later acquired by VMware. Where I learned to test in the open.
Started in 2012 at CleanTelligent, manually testing tickets and teaching myself to automate. The full history is on the resume.
· how I work
The advantage isn't AI.
It's managing AI well.
Everyone has access to the same models. What changes the output is the surrounding system — clear context, opinionated architecture, real infrastructure. That's the differentiator.
- 01
AI does the typing. I do the thinking.
Every project has a CLAUDE.md that captures architecture, conventions, and live work-in-progress. Tome's is ~25k lines. I pair with Opus 4.7 to compress days of typing into hours of decisions.
- 02
Ship small, ship often, ship to prod.
Tome moved from v0.11 to v0.12.3 in a month — small versions, real deploys, real users. My homelab CI/CD agents run on Raspberry Pis I built myself. Most projects deploy on push.
- 03
Own the whole stack — including the rack.
Production for me means a Proxmox cluster in my closet running Tome, Pi build agents, and a Home Assistant bridge. Trusting my own ops more than someone else's marketing copy is the cheat code.
tome / CLAUDE.md
# CLAUDE.md (excerpt) — Tome
## Architecture
- Server: Node + Express + TS, Supabase PG
- Client: Flutter (iOS / Android / web)
- Reader: paginated EPUB, synced progress
## Conventions
- Endpoints return { data, error } — never raw
- Migrations live in /supabase/migrations
- Never trust client-supplied book metadata
## What I'm working on right now
- v0.12.3 — paginated reader, x-device sync
- Up next: audiobook chapter scrubbingexcerpt · trimmed for fit~25,000 lines total
· let's talk
Hiring, collaborating, or curious about the workflow.
I'm most useful when I get to own a problem end-to-end — QA strategy, the platform itself, the operational layer underneath. If that sounds like something your team needs, let's talk.
© Chris Arroyo · built with the same workflow on display
chrisarroyo.dev