Khivi

Engineering Outcomes

Every line of code is a liability. I build systems where the complexity lives in the business domain, not the technology. Simple code, clear interfaces, boring technology that your team can master — that's what ships reliably. AI is no exception: it needs guardrails and a clear return, not just output. Every diff answers the same question — is this smaller, and is it worth the complexity it adds? What evolved isn't the bar; it's where the effort goes — less time writing code, more deciding which code earns its place.

I design for change, not perfection. Customers invest in outcomes, not technology. I make architectural decisions that keep options open, delay irreversible choices to the last responsible moment, and keep the cost of iteration low. When your assumptions change — and they will — your system should adapt gracefully.

Execution Culture

Strategy without execution is a slide deck. I build execution cultures — where sprints actually work, PRs get real reviews, master is always deployable, and deployments are zero-touch. I've codified this into what I call The KHIVI Test: ten questions that tell you if your engineering org is actually shipping or just busy. AI-written code clears the same gate as everyone's — tests, automated review, security scan, master always green. The model brings speed; the gate brings trust. Our cadence evolved with it — smaller diffs, shipped more often, each one cheap to generate, review, and revert.

I believe in small experiments over grand plans, fast feedback over perfect specs, and teams that own their quality end-to-end. Product, engineering, and business aren't separate functions — they're one team solving one problem.

Mentorship

The best engineers I've worked with weren't born great — they were mentored well. I challenge engineers to come up with their own solutions, guide them through the tradeoffs, and create environments where they can take ownership. In the AI era that means teaching them to interrogate the model — review its code like any other, ask "why is this better?", keep the judgment human. The craft itself evolved: juniors once grew by writing the loop; now they grow by judging the one the model wrote — which demands taste sooner. I wrote Culture Seeds to capture the patterns that build great teams: how to hire, how to set standards, how to create a culture where quality is the path of least resistance.

Just released a new edition of Culture Seeds — updated for engineering in the AI era, where your tools can generate code but can't yet think about systems.

Building with AI

Call it vibe coding if you like — but I treat AI like any other part of the stack: deliberately. The model writes the code; I own the system. Tests stay non-negotiable, reviews stay real, and nothing ships on vibes alone. AI made me faster, not looser.

So I built a platform around it, not a habit of pasting prompts: dozens of skills that encode my workflows, specialized review and security agents, multi-agent orchestration, file-based memory that persists what matters, and a token-proxy CLI that cut my usage by more than half. It's the same instinct as the rest of my work — optimize the system, not the parts. More on the how in my Substack and the AI-era edition of Culture Seeds.

Philosophy

Journey

Technology

Testimonials

Connect

Building something? I can help.

Architecture reviews, hands-on coding, team building, testing strategy, production debugging, or figuring out what to build next — I've done it all and I'm still doing it.

Email Me