Spec-driven development
Requirements written to be testable, not to be read. Delta specs, acceptance criteria, and a traceable link from every requirement to the test that proves it.
Hanoi, Vietnam
AI systems that stay aligned with the work they are meant to do.
Ai Pro is a small engineering studio. We build DrawSI — a tool that turns architecture diagrams into verifiable specifications — and the agent workflows that keep AI-written code on target instead of drifting.
What we do
Three things, done properly.
Most AI engineering fails at the same place: nobody can tell whether the output is correct. We work on that problem.
AI agent workflows
Multi-agent setups with a governance layer on top: a charter, recorded decisions, and hard scope limits. Built so every change an agent makes can be reviewed and audited.
Developer tooling
TypeScript, monorepo build graphs, browser automation and Chrome extensions — tooling made for engineering teams that ship, not demos.
Product
DrawSI
Diagrams and documentation drift apart the moment someone edits one and forgets the other. DrawSI treats a diagram as the source of truth and generates specifications from it — so the document cannot silently fall out of date.
The result is a document you can verify, not just read.
How we work
Decide, specify, verify.
01
Charter
We write down what the product is for, and what it is explicitly not for. Everything later is measured against that one page.
02
Specify
Each change starts as a short proposal with a stated goal. No work begins until the goal is written and agreed.
03
Verify
Every requirement gets a test. A change is done when the test passes and the specification has been updated — not when the code compiles.
Working on something that needs to hold up?
Tell us what you are building and where it currently breaks down. We reply to every serious enquiry.