miyamotoai
Joshua Tseng · AI Systems Architect
[00%]
AI Systems Architect · Consultant

AI systems that ship.

Automation, infrastructure, and applied AI for businesses that need to move faster than the competition.

Scroll to explore
Claude APIn8n AutomationDockerPostgresReactTypeScriptPythonCaddyLinuxGmail APIHetzner VPSSelf-hosted AILLM Pipelines Claude APIn8n AutomationDockerPostgresReactTypeScriptPythonCaddyLinuxGmail APIHetzner VPSSelf-hosted AILLM Pipelines
ship
build
deploy
Manifesto

Most AI projects die in Notion.

They get workshopped in meetings, demoed in sandboxes, and shelved before they touch a real workflow.

I build things that actually run. In production. On your data. In your environment.

No vendor lock-in. No black box. No waiting for an engineering team to prioritize your ticket.

Just fast, private, working AI — deployed in 24 hours or less.

Philosophy
Ship before you optimize. Real data beats perfect architecture every time.
Audience
Operators tired of waiting. Companies that can't afford to be slow.
Approach
Self-hosted by default. Open infrastructure. You own everything.
Services
01 — Automation
Intelligent Automation

End-to-end AI pipelines replacing manual workflows. Gmail to Postgres, Claude API to n8n — built and running in production.

02 — Infrastructure
Self-Hosted Infrastructure

Private AI stacks on your own server. VPS, Docker, Caddy, Postgres. Your data never leaves your environment.

03 — Product
AI Product Portals

Client-facing interfaces in React + TypeScript. Prompt engineers, dashboards, ops assistants. Production-ready.

Selected Work

Featured
projects.

002 / projects
Project 001 · Bay Area, CA · 2024

Highland
Build

Intelligent email operations for a Bay Area general contractor. Priority scoring engine, automated draft generation, Claude API processing, Postgres audit logging.

80%
email overhead eliminated — week one
Gmail APIClaude APIPostgresn8nPython
Request case study →
Project 002 · Self-hosted · Live

miyamotoai
Portal

4-agent AI platform — prompt generator, visual whiteboard, UI/UX critique agent, infrastructure ops assistant. Deployed on Hetzner VPS. Private, fast, zero cloud dependency.

4
AI agents running in production simultaneously
ReactTypeScriptDockerClaude APICaddyHetzner
View live portal ↗
80%
Email overhead eliminated — Highland Build, week one
4
AI agents running in production simultaneously
24h
Typical concept-to-deployed pipeline cycle
Uptime on self-hosted infrastructure since launch
How I work
Strategy-led.
Execution-first.

No sprints, no stand-ups. I embed into your workflow, identify the highest-leverage AI intervention, and ship it fast.

01
Audit

Map your existing workflows. Identify where time is being destroyed. Find the 20% of processes causing 80% of friction.

02
Architect

Design the minimal viable AI system — one that solves the problem without creating three new ones. Private by default.

03
Deploy

Ship to production in 24–72 hours. Running on your server. Observable, auditable, maintainable. No vendor dependencies.

04
Iterate

Tune against real data. Expand the system as confidence builds. Leave you with a codebase you actually own.

Get in touch
Let's build something.
Available for consulting & full-time roles