I build the AI that answers your customers, on the phone and in your inbox. It gets the answer right, and it hands you the ones it shouldn't touch. Already have one that keeps getting it wrong? I fix those too.
It's almost never the AI itself. It's what sits around it: what it can look up, and what it does when the lookup fails. That's what I get called in to fix.
The second it hits a question it can't answer, it invents one and says it with total confidence.
Ask "is it delayed?" and it no longer knows what you meant two messages ago.
Your order system goes down. The AI never mentions it. It just guesses and keeps going.
The customers who actually need you get a confident wrong answer instead.
One bad loop, or one person spamming it, and a month of budget is gone in a day.
It forgets everything each time it restarts, so every fix quietly breaks something else.
A traveler asks about their flight and gets a real answer. Gate changes, delays, a connection they've already missed. English and Turkish, reading live flight data. If it doesn't have a flight number, it says so.
Built for a client getting hundreds of messages a day. It sorts the real questions out of the noise, and it remembers every customer from one conversation to the next. Anything it's unsure about goes to a person.
Spanish-language support. Customer messages all land in one shared inbox, and whatever the team replies there goes straight back to the customer on Telegram.
Two strangers trade, and the money sits locked in the contract until both sides do their part. About 2,000 lines of code, with 68 out of 68 tests passing. An outside security firm reviewed it, and all 8 issues they found were fixed before launch.
I work out what's actually broken before I quote you a price.
Clear price, clear timeline. If I haven't built it before, I tell you.
I design around those six problems from day one.
The code is documented and you get a step-by-step guide, so your team can run it without me.
Yes. Most of my work starts from nothing. You tell me which calls or messages are eating your day, and I build the thing that handles them.
Then I tell you that, and you keep the write-up. Sometimes the honest answer is to rebuild it a different way.
The first look usually takes a few days. What gets fixed, and what it costs, we work out from there.
Either. Small fixes work best fixed-price; larger work runs hourly or in milestones.
Claude and GPT APIs, RAG, MCP, multi-agent orchestration, voice agents, Python and FastAPI, Next.js, vector databases, and the webhook and integration plumbing that ties it together.
Yes. The reliability work is the same either way.
The AI I build is running right now, talking to real customers and moving real money. You get documented code, plus a guide your team can actually follow. On price, I'll give you a straight answer. The tools I work in, if that matters to you:
Tell me what's happening. I'll tell you what it takes to fix and what it costs. If it isn't worth doing, I'll say that too.
Usually replies within a day · Taipei (GMT+8)