How I work

Most clients have never seen a delivery process. Showing one builds trust immediately. AI is part of it — as a tool that speeds up delivery, not as a slogan. I review the architecture and the code by hand.

1. 1
   Discovery — I listen to the people doing the work.
2. 2
   Business analysis — I map the real process and the bottlenecks before any code.
3. 3
   Architecture — A design your team can maintain — not a black box.
4. 4
   AI-assisted development — Agents draft; I review architecture and code manually.
5. 5
   Testing — Automated tests alongside the feature, not after.
6. 6
   Deployment — CI/CD ships it; nothing built on the server.
7. 7
   Support — Observability + handover so it keeps running.

What you can count on

• Fixed scope first. I write down what's in and what's out before any code, so price and timeline don't drift.
• Availability. Predictable check-ins — you're never wondering where the project is.
• Testing. Automated tests alongside the feature, not "we'll add tests later".
• Handoff. Docs, deploy steps, and observability so your team can run it without me.

The tools behind it

AI accelerates the boring parts; the consequential parts stay human-reviewed.

• Pi.dev — Agent harness
• Claude Code / OpenCode — Coding agents
• Windsurf — IDE-embedded agent
• Context7 — Live library docs
• Serena — Symbol-level code navigation
• GitHub Copilot — Inline completion

See it in practice

The process above is exactly how the case studies were built. Each one follows Problem → Analysis → Architecture → Implementation → Result → Lessons.