Hi, I'm Dovid Gefen

Background

8+ years building and managing cloud infrastructure — AWS, Azure, and GCP. I've set up CI/CD pipelines, migrated production workloads, locked down security, built monitoring from scratch, and cleaned up the kind of infrastructure that makes you wince when you open the console.

The pattern I kept seeing: companies where DevOps was nobody's actual job. The CTO or a senior dev was handling it on top of everything else, and the infrastructure reflected that — brittle, undocumented, held together by one person's memory.

Why I Built the AI Harness Approach

The classic consulting model has a problem: I come in, fix everything, leave, and six months later the infrastructure drifts back because the team didn't have a way to maintain what I built. The knowledge walked out the door with me.

The DevOps AI Harness is my answer to that. Instead of just fixing the infrastructure, I build a Claude Code setup around it — context files that encode how the systems actually work, CLI tools wired into the real environment, runbooks built into the workflow. The harness gets built as part of the engagement because it also helps me work more efficiently. When I leave, the team inherits an AI that understands their stack, not some generic chatbot.

The difference between a useful AI setup and a useless one is what you put in front of it. That's where 8 years of infrastructure experience matters — knowing what context the AI needs, how to structure it, and what operational knowledge to encode so the team can actually rely on it.

How I Work

Engagements run 2–3 months. I start with a small paid discovery ($800) so both sides know what the work actually is before committing to a buildout.

After an initial ramp-up period, I work mostly independently — minimal meetings, no joining your daily standups or Slack channels. I'm available if an urgent DevOps issue comes up, but otherwise I keep the friction on your team to a minimum.

At the end of the engagement, I train your CTO or lead dev on the new system — though it's straightforward and doesn't require much to get the most out of. The project is only complete once my changes have been tested thoroughly and you're fully satisfied with the work.

I'm direct about what I find, even when it's not what you want to hear. If your infrastructure is fine and you don't need me, I'll tell you that during discovery. If something is a bigger mess than expected, you'll know that too — before the buildout starts, not after.