How I work

AI velocity.
Engineering discipline.

I’m an AI-first developer. Coding agents write a large share of my first drafts, connected through MCP to the tools where real work lives. That’s the velocity. The discipline is what happens next: every architectural decision, every library choice, and every line that ships goes through twelve years of production judgment. Mine.

Agents

Claude Code
Codex
Gemini CLI
Cursor

Connected to

Jira
Figma
GitHub
Firebase

The difference

Not a vibe coder. An engineer with a faster pen.

Vibe coding

Accepts whatever the model outputs and moves on.
Lets the AI pick libraries and patterns on the fly.
Prompt, paste, pray. Architecture emerges by accident.
Can’t explain the code when review gets hard.

How I work

Reviews, corrects and re-architects the output with 12+ years of production judgment.
I decide the stack and steer agents toward the right library — choices that stay future-proof, with fewer bugs and less repair work later.
Architecture comes first. Agents implement to my spec — module boundaries, interfaces and state management are decided before generation starts.
I own every line that ships — and can defend any of them in a code review.

The pipeline

Four steps, one owner.

Decide

Architecture, tech stack, module boundaries, interfaces. The decisions that make or break a codebase are made by me, before any code is generated.

Delegate

Coding agents — Claude Code, Codex, Gemini — implement to spec, connected through MCP to the real tools: Jira tickets, Figma designs, GitHub, Firebase.

Review

Line-by-line. I track what the agents used, correct wrong turns, and re-architect where needed. AI accelerates; it doesn’t get a vote.

Ship

Tests, CI/CD, store release, production monitoring. The same delivery discipline I’ve applied for 12 years — just reached a lot faster.

This isn’t theory. V-Assist — a production drone app — used AI tooling across its full development lifecycle. At Avetics I embedded Claude Code, Gemini CLI and Cursor into a 6-person team’s daily workflow. Football Predictor 2026 and Study Quran were shipped solo, at a pace that surprised everyone except the workflow. The results are on the projects page.

Want this velocity pointed at your product?

Build your product →