Skip to content

Harness engineering for AI agents · Code audits · Code takeovers

Your code output has exploded.You can no longer review it all.

A developer for twenty years, I help development teams move to AI agents, or make better use of them, without losing control of their code.

Four ways I can step in

For teams producing more code with AI agents than they can review.

Harness engineering: speeding up with AI agents while keeping control of the code

Harness engineering builds the environment AI agents work in: the context they receive, the rules they follow, and automated checks on every commit and in CI. Whether the team is starting with agents or already uses them, I set it up with the team, for example with Claude Code, and we put it to the test together on your real tasks. Everything stays in your repositories, documented.

Taking stock of your code, and fixing what is urgent

I read the code, the architecture, the tests and the CI, then how the team works day to day. You get a written assessment with the priorities in order, and I can handle the most urgent ones right away: fixes, updates, missing tests. First findings come within a few days.

Moving code the team no longer dares to change

I take the code where it is stuck: what breaks with every change, what slows the application down, what nobody understands anymore. The goal is for the team to ship fast again without fear of breaking everything, whether the code came from agents or from a framework that is no longer maintained.

Topics I work on

  • Architecture checks in CI
  • Taking over agent-written code
  • Laravel upgrades
  • Migration to Go
  • Performance and slow queries
  • Integrations and APIs
  • Team leadership

Your situation fits none of these? Describe it, and we set the scope together.

Let's discuss your situation →

Selected engagements

Recent engagements, in different contexts.

Go · SaaS platforms

Microservices and business rules

Moving Laravel components to Go on a data processing platform, then leave-tracking history and carry-over rules on another engagement.

Moving code the team no longer dares to change

Olivier Drot

Behind notmybug, Olivier Drot.

Twenty years of software development, five of them as lead developer, first in industry (EADS, Airbus, Naval Group), then on SaaS platforms and business applications. I use AI agents every day, and I build Synapse, my lab: an agent orchestrator where I test what I later set up for clients.

Discover my background →

Frequently asked questions

Who gets access to our code?

I work in your repositories with the access you grant me. When I use an AI agent on your code, it goes through an account you control.

Do you sign an NDA?

Yes, before any access to the code.

Do we need to use AI agents already?

No. Audits, code takeovers and the CTO role do not depend on agents. Harness engineering is for teams that use them or want to bring them in.

Which tools do you use with agents?

Your team’s: Claude Code, Codex or Gemini. If another tool suits your use better, I tell you why, and the decision is yours.

What stays after you leave?

Everything that was set up stays in your repositories: automated checks, agent rules, documentation. Nothing depends on me to keep working.

What would you like to move forward?

Your codebase, your AI workflows or a specific difficulty: a few lines are enough. Scope gets clarified in an initial conversation.

Or write directly to contact [at] notmybug [dot] dev

Your data: see the privacy policy

Remote engagements · French and English