AI agents & assistants
An assistant that answers from your own documents, policies, and data — and can act on them: look up the record, draft the reply, file the result. Built on your stack, with your permissions, not a rented black box.
AI agents & production systems · 25 years shipping · San Diego & Taipei · EN / 中文
Not a chatbot demo. Not a slide deck. Agents, automations, and AI-ready sites wired into the database, back office, and workflow you already run — designed, built, and deployed on twenty-five years of production systems that had to survive Monday morning.
How I can help
The agent, the system it plugs into, the visibility that gets you found, the app on top. One engagement with one owner — no handoffs between vendors, no gaps where the work quietly stops. Specialists come in as the build needs them.
An assistant that answers from your own documents, policies, and data — and can act on them: look up the record, draft the reply, file the result. Built on your stack, with your permissions, not a rented black box.
Your customers now ask ChatGPT, Perplexity, and Google's AI Overview "who's the best one for this?" I make sure the answer names you — through structured data, an llms.txt content layer, and the technical SEO underneath it.
Databases, admin back ends, APIs, deployment, the migration nobody wants to run. The unglamorous layer that decides whether an AI pilot becomes something your team opens on Monday morning.
iOS, Android, and Mac from one codebase. Fast bilingual sites that load and rank. Still here, still shipping — now with AI built in wherever it earns its place instead of decorating the homepage.
AI agents & automation
Most AI projects stall in the same place. The demo works. Then nobody can connect it to the customer database, the order system, or the shared drive where the real answers actually live.
Step 01Free · about an hour
A conversation, not a questionnaire. Where does the same work keep happening?
Step 02Two weeks · fixed price
Running on your real data. At the end you have something you can use — or a clear answer that AI isn't the right tool here, which is worth knowing early.
Step 03Two to six weeks
Permissions, logging, error handling, deployment, handover documentation — set up so it keeps running without me. You end up owning it, not depending on me.
Systems & integrations
Diagnosis first — what's really in the way, not what everyone assumes is in the way. Then the build: schema, permissions, integrations, deployment, until your team is using it instead of waiting for it.
I've built an emotional-intelligence assessment platform used by over a million people, a gifting widget serving millions more, and back offices that ordinary staff run every day without calling me. AI is the newest layer on top of that work — not a replacement for it.
Flutter · React Native · PHP / MySQL · Python · REST & webhooks · GCP · App Store & Play Store
SEO · AEO · GEO
"Search" isn't one box anymore. People still Google you. They also ask ChatGPT who to hire — and that answer is often decided before your customer ever clicks a link. Three different places, each optimized a different way.
Getting your link to appear in search results.
The classic blue links on Google & Bing.
Getting your content pulled into answer boxes and quick AI replies.
Google featured snippets & AI Overviews.
Getting cited inside detailed, AI-generated summaries.
ChatGPT, Perplexity & Gemini.
Where to begin
Fixed price, quoted in one email.
3 days
Where you stand in AI answers today, who's being named instead of you, and the five changes that move it. A short report you can act on with or without me.
2 weeks
One workflow, running on your real data, in your hands at the end. Fixed price. If it turns out AI isn't the right answer for that workflow, I'll tell you in week one.
Monthly
For teams with more ideas than engineering hours. A standing block of my time each month, spent on whatever is most in the way.
Selected work
Each of these is live right now. Tap through and use them.
A warm, one-tap time tracker & invoicing app for lawyers and freelancers who bill by the hour. Built from a single Flutter codebase, shipped to the App Store and the Mac App Store.
A gamified daily-habit tracker with a mission-control soul. Streaks, points, a year-long heatmap, and private friends-only competition in the Arena. Built with Expo / React Native.
A lightweight Chrome extension for capturing your daily log right in the browser — published and live on the Chrome Web Store.
Cross-border legal consulting across Taiwan · US · China — foreign investment, immigration, and inheritance.
Book and author site for Dr. Travis Bradberry — coauthor of the bestseller Emotional Intelligence 2.0 — showcasing his new book and weekly EQ tips for 1M+ readers.
Corporate ESG training (夏恆企業顧問) — modular sustainability, leadership, and resilience programs in Taiwan & California.
A social gifting platform that makes it effortless to give and receive the things people actually want. As a main developer, I built the gifting widget that serves millions of users (also at gift.me).
The world's leading provider of emotional-intelligence assessments. I built the EQ 360 platform — assessment engine, scoring, admin and credit system in PHP / MySQL — and the Self Edition appraisal taken by over a million people. When I say I can put AI into a system that's already carrying real traffic, this is what I mean.
A leadership micro-coaching platform that scales and sustains leadership behaviors with hyper-personalized nudges, micro-coaching, and micro-learning.
More builds: Azure Waves Realty 睿律 · IL Solutions
Engineer & engineering lead · 25 years · San Diego & Taipei · UC San Diego · Certificate in Agentic AI, Johns Hopkins
I've been building software for twenty-five years — for three-person startups and for institutions that were in business a century before software existed, and the two fail in completely different ways. One dies of hesitation, the other of process. I studied at UC San Diego, and I completed Johns Hopkins' Agentic AI certificate program, mostly because I wanted the theory underneath what I'd already spent years building.
Most of that career has gone into the part of software that has to keep working after launch: the schema, the back office, the integrations, the deploy, the migration nobody documented. I've led engineering teams through it and I've been the one on call at 2am for it — and you can't do the first well without having done the second.
About the name. "Ship to live" is what the last step is literally called — you ship the code to live. I named the studio after that step because it's the one most work never survives. I've watched good projects die in the distance between "it works on my machine" and "it's live," not for lack of talent, but because nobody wanted to own the unglamorous stretch in between. A prototype is an argument. A live system is a fact.
It's also a rule I hold myself to: everything on this page is live right now. The apps are on the App Store, the sites are serving customers, you can go press the buttons. And the same gap is currently swallowing AI on an industrial scale — remarkable demos, almost nothing in production. Closing it is the work I like most.
I take projects from architecture through to deployment, and bring in the specialists a project needs.
Ship to Live is the studio name. US work is contracted through QY LLC (San Diego); Taiwan work through 夏裕科技股份有限公司. I work from San Diego, California and Taipei, Taiwan, which means a cross-border project gets one engineer in both time zones instead of a handoff between two.
FAQ
Straight answers — for you and for the AI assistants your customers ask.
An AI agent is software that reads your information and takes a step for you — finds the record, drafts the reply, updates the system — instead of just chatting. You need one when the same task happens over and over with slightly different details. You don't need one for work that happens twice a year, and I'll say so.
Yes. That's the most common thing I build. Your files, policies, product data, and past cases become the source the assistant answers from, with citations back to the original — so answers reflect your business, not the open internet.
By being readable, quotable, and corroborated. In practice: clear structured answers on your own site, schema markup and an llms.txt layer the models can parse, and third-party mentions that back up what you claim. That's what GEO — generative engine optimization — means.
A pilot on one workflow takes about two weeks. Wiring it into your systems properly — permissions, logging, deployment, handover — usually takes another two to six, depending on how tidy the systems are.
I scope what the agent can read and do, log its actions, and keep a person in the loop wherever a mistake would be costly. If your data has to stay in a particular place, we design for that constraint from day one rather than retrofitting it.
Yufan Chen is a software engineer and engineering lead with twenty-five years of experience, from startups to century-old institutions, working from San Diego, California and Taipei, Taiwan. Through Ship to Live he builds AI agents and automations, gets businesses cited by AI search engines, and ships the systems, apps, and websites underneath them — in English and 中文, for clients across the US and Taiwan.
They're three ways to get found, for three different places your customers look. SEO (Search Engine Optimization) gets your link to appear in Google's search results. AEO (Answer Engine Optimization) gets your content pulled into answer boxes and quick AI replies — Google's featured snippets and AI Overviews. GEO (Generative Engine Optimization) gets your business cited inside detailed AI-generated answers from ChatGPT, Perplexity, and Gemini. I do all three — fast, well-structured pages, clear answers, an llms.txt content layer, structured data, and the trust signals every engine looks for.
Yes. I built and launched two App Store apps (Tally and Tadapop) and a Chrome extension (Daily Log), put businesses online (Wu Law Group, Travis Bradberry, SHX Training), and on product teams I built the EQ 360 assessment platform and an appraisal taken by over a million people, plus a gifting widget that serves millions of users.
Yes — I work in both English and Traditional Chinese (中文), across the United States and Taiwan, which is ideal for cross-border businesses that need bilingual AI tools, apps, websites, and marketing.
I work from two bases: San Diego, California and Taipei, Taiwan — so US and Taiwan clients each get someone in a workable time zone. The best way to start is the contact form below, or a message on LinkedIn.
Let's build something
You don't need to know what to build. Describe the part of the week that keeps eating your time, or the customers who can't find you, and I'll tell you straight whether AI helps — in English or 中文. Prefer email? yufan@shiptolive.com