I'm Andy Wang, an AI engineer based in Xi'an (UTC+8) — more than two decades building web applications, primarily in Ruby on Rails since 2008, remote with US and European teams since 2011. Right now: production AI workflows at Adwave, and my own independent products, chiefly HexagramCast.
Now / what I'm working on
Key contributions to the production AI workflow behind a Rails platform for video ad generation and campaign management — AI tasks, background jobs, external services, and deterministic media generation cooperating as one process, plus agentic editing and the operational tooling around it.
My own product: an English-first, source-backed I Ching reflection tool with web, API, and browser-extension surfaces. Traceable casting, verifiable computation, and a deliberate non-fortune-telling stance. English, 简体中文, 繁體中文.
Long-term ownership of business-critical Rails codebases — payments, scheduling, marketplaces, operational admin and reporting — plus LLM features on RubyLLM and Hotwire-driven portals that feel native without leaving the monolith.
About
I've been a Rails developer for about as long as Rails has been a thing people did for a living — and a remote one for almost as long. Most of what I do is the unglamorous, careful work of turning business rules into systems that keep working while they change.
I started writing code professionally in 2005, building projects for Japanese customers in VBS, C#, Java, and HTML. In 2008 I went all-in on Ruby on Rails at Idapted, leading a small team on a 1-on-1 English-learning platform. Since 2011 I've worked remotely — first at Intridea in the D.C. area on more than ten client projects, then a long string of contracts including Spire Health and imToken, and a multi-year run owning the codebase for a Stripe-powered swimming-lessons platform.
Since January 2025 I've made key contributions to Adwave's production AI workflow — a Rails platform for video ad generation and campaign management, where AI tasks, background jobs, external services, and deterministic media generation all have to cooperate. My part is the product system around the model: agentic editing, structured state, observability, testing, and the browser-based QA that keeps it honest. It's a client product, and the interesting work is what it takes to make something like that dependable.
Alongside it I've worked independently with international clients since 2020, on business-critical workflows rather than isolated websites: a Stripe-powered platform for swimming products and lessons with instructor availability and scheduling rules, a configurable CMS combining scheduling and payments for paid online services, room and meeting operations across a large building, a specialised vendor marketplace, a multi-party private-investment platform, and a publishing workflow spanning Google Docs, Medium, HubSpot, and WordPress.
That same practice runs my own product work, chiefly HexagramCast — an English-first, source-backed I Ching reflection tool with web, API, and browser-extension surfaces, built on traceable casting and verifiable computation, with a deliberate non-fortune-telling stance. It's where the whole path stays visible: framing the problem, building it, shipping it, charging for it, and running it. Around it I keep exploring niche SaaS and public-data products as a practical way to find problems worth solving.
Outside work I read, walk, and translate Buddhist sutras — chapters of the Diamond Sutra and the Heart Sutra — for my own practice. A product engineer by day; a life-long learner by night.
What I do
Models put to work inside real products, not demos: agentic workflows, structured outputs and tool calls, background jobs and async state, retries and fallbacks, deterministic steps where determinism matters, plus the operational layer — evals, observability, cost control, user feedback. On the Rails side, RubyLLM and its ecosystem are home ground.
Senior-level Rails engineering, end to end. Greenfield products, long-term ownership of mature codebases, and the awkward middle: the app that worked fine until it didn't.
Stripe integrations with real business logic — subscriptions, marketplace splits, refunds, dunning. Several years on a swimming-lessons marketplace combining coach availability, complex course rules, payments, and admin reporting.
Systems where the hard part is roles, permissions, documents, and state: private-investment workflows, vendor marketplaces, room and meeting operations across a large building, configurable customer-facing portals.
Rails content systems with sprawling third-party reach — publishing into Google Docs, Medium, HubSpot, and WordPress. Clean REST & GraphQL APIs, plus the messy half: webhooks, retries, and assorted vendors who consider documentation optional.
React for the parts that genuinely benefit from React; Hotwire/Turbo for the parts that don't. I don't believe in framework loyalty — I believe in shipping the thing.
A short timeline
Stack
Words
Andy is very intelligent and has a strong understanding of web development principles. I've always been comfortable handing him a project and knowing that he'll be able to solve even the most difficult problems.
Andy is an experienced software engineering professional that excelled at tackling every single challenge thrown at him. I had witnessed him pick up several different codebases and be instantly productive.
A superstar with a finely honed ability to deliver well-tested, well-executed code. Consistently amazed with his ability to think creatively to solve hard problems with the resources on hand — usually while juggling two or more projects at once.
Any kind of bug or feature he took on was done quickly and done well. He communicated very clearly when he needed clarification. I never needed to worry that something would get done.
I'm most useful where there are real users, non-trivial workflows, and a codebase that must keep running while it changes. Open to conversations about AI workflow integration, Rails systems, senior product engineering, and end-to-end delivery — usually one or two engagements at a time. UTC+8, but well practiced at meeting US/EU teams in the middle.