Custom builds
We design, build and operate custom AI and product systems for businesses that cannot afford downtime. When the work is done, you own every part of it.
What we build
01
Voice agents, LLM assistants, and the evaluation infrastructure that keeps them honest.
Typical work
02
Full-stack product engineering from architecture to operations: event-driven backends, modern web frontends, and infrastructure that scales predictably.
Typical work
03
Deep Shopify and LINE platform work: storefronts, admin apps, integrations, and the messaging rails Thai commerce runs on.
Typical work
How we work
Five stages. Each one ends with something you can see, and nothing moves on until you have seen it.
We learn the business before the tech: who uses it, what breaks today, and what success looks like in numbers you already track.
You get A written brief and scope
We pick the simplest design that will hold, write down the tradeoffs, and agree on what we will not build.
You get Architecture notes and a phase plan
Short cycles with working software at the end of each one, on a staging environment you can open any time.
You get Running software, every cycle
We run it in production with monitoring, alerts and runbooks, and fix what real use reveals.
You get Monitoring, alerts and runbooks
Your team, or the next team, gets everything: the code, the infrastructure, the docs and a recorded walkthrough.
You get A complete handover kit
What you own
At the end of an engagement the system belongs to you, in a form that never depends on any one person. Including us.
The full source in your own repository, with its history.
Every server, queue and secret reference, rebuildable from files.
Architecture, decisions, and how the pieces fit together.
Step by step for deploys, rollbacks, incidents and routine care.
A video walkthrough your next hire can watch on day one.
Sample deliverables
Three excerpts from our own builds, with some details redacted. Custom systems we build for clients are handed over in the same shape: decisions written down, tests with a receipt, and a runbook your team can follow.
Solid bars mark details we redacted.
Q5 Lead capture on the website
Status: accepted, 2026-10-07
Decision: contact form -> one serverless function -> email and an auto-reply. No database of personal data.
Why: we only need the message. Data we never store cannot leak and has nothing to purge.
Trade-off: the inbox is the record; retention follows the email rule in the privacy notice.
Approved by:
Every significant choice is written down with its reason and its trade-off, so the next engineer never has to guess.
$ npx playwright test
pages: 12 routes × EN, TH × light, dark
devices: Desktop Chrome 1440 · Pixel 7
each page: axe WCAG 2.2 AA, no console errors, no sideways scroll, one h1
flows: contact form, caption tracks, mobile menu focus
commit:
212 passed (6.4m)
What was tested, on which devices and themes, and the result, in a form your team can rerun.
Rollback
$ npx wrangler deployments list
$ npx wrangler rollback
Back to the previous version at once, no build. Secrets and bindings do not roll back: test the contact form again.
Rotate a secret
1. Create the new credential; keep the old one valid
2. $ npx wrangler secret put <NAME>
3. Smoke test the form on staging, then production
4. Revoke the old credential at the provider
5. Log the date and reason, never the value
Account:
Deploy, rollback and secret rotation written as steps your own team can follow without us.
Ways to work together
Every engagement is quoted in writing after a scoping call, so you know the price before any work starts. Most work takes one of these three shapes.
A defined system with a clear finish line.
Steady ownership of your product, month after month.
Senior guidance for the team you already have.
Tell us what you are building and where it hurts. We will reply with honest next steps, even if that means a smaller project.
Running a clinic? See our product: Clinic AI Receptionist