I lead AI transformation for a global SaaS platform, helping 100+ engineers adapt to a world where AI performs much of the execution and humans remain responsible for the outcome.
I build AI-powered systems that investigate problems, reproduce bugs, write and review code, validate changes, and automate operational workflows. The goal is not to replace engineers. It’s to remove the repetitive work that prevents them from doing real engineering.
This is not AI experimentation. It’s the operational redesign of engineering work. This blog is where I write what I learn doing it.
What I actually do#
Most AI initiatives begin with a model, a demo, or a list of tools. Mine begin with work that should no longer be performed the same way: a bug report that eats hours of investigation, a support escalation that bounces between teams, a workflow that depends on someone manually collecting context, a production process that exists because nobody has redesigned it yet.
I work with the people closest to the problem and learn how the workflow actually operates — the systems involved, the hidden decisions, the exceptions, the ownership gaps, the reasons previous attempts failed. Then I build the system: agents, automations, code-generation workflows, evaluation pipelines, approval gates, integrations, observability, and the infrastructure that makes all of it reliable.
The work doesn’t stop when the demo succeeds. The system has to survive production, earn trust, produce measurable value, and fit how the organization operates. Then the one-off solution has to become something reusable. That’s the difference between an AI experiment and organizational transformation.
How I think about engineers and AI#
I don’t believe the future of engineering is humans competing with AI over who types code faster. AI will increasingly perform the implementation. The engineer’s value moves elsewhere: defining problems, understanding systems, designing architecture, making trade-offs, evaluating generated work, identifying risk, and taking ownership of production.
The agent can produce the code. The engineer must know whether the code should exist, whether it solves the right problem, whether the design is sound, and whether the result is safe to ship.
AI changes the execution model. It does not remove human accountability.
How I work#
My approach looks like a forward-deployed model inside my own organization: get close to the teams experiencing the problem, don’t wait for perfect requirements, build the first production solution, measure whether it changes the outcome, and turn the learning into a reusable capability.
The process is usually:
- Find a meaningful operational or engineering problem.
- Understand how the work is actually performed.
- Split the decisions that need humans from the execution that doesn’t.
- Build the smallest system capable of producing real value.
- Put it into production.
- Add the controls required for trust, safety, observability, and accountability.
- Measure adoption and business impact.
- Turn the pattern into infrastructure, a platform capability, or an operating model.
A successful project is not a compelling prototype. It’s a system people actually use, that survives production, and that materially changes how work gets done.
How I got here#
I’ve been taking things apart since I was eight. Electronics on my bedroom floor, things my parents probably wanted back in one piece. By my teens I was deep in IRC channels, writing scripts, hustling into communities that didn’t care how old you were as long as you could keep up.
I’ve been in tech professionally since I was 22. IT support, NOC engineering, freelance dev, full-stack, automation, DevOps, architecture, management. I didn’t plan a career path. I just kept saying yes to whatever scared me most.
The decade-long bet was joining a startup as employee #5. No DevOps. No infrastructure. Barely a product. I built it all. The test framework, the CI/CD, the first deployment pipeline, the first monitoring. When we needed an offshore team, I flew to Ukraine, hired the people, made it work across timezones.
That startup is now the global SaaS platform where I work today. As it grew, my role moved from building systems myself to building the platforms, teams, and capabilities that let others move faster. When AI agents became viable, I treated them the way I’d treated infrastructure years earlier: not as a trend to observe — as a capability to operationalize.
Today my focus is no longer DevOps. It’s redesigning how engineering organizations operate when intelligent systems can perform large parts of the work.
Why read me#
Five things you’ll get here, consistently:
- Production-grounded takes. The systems I write about touch real workflows, real codebases, real incidents, real budgets. If an idea doesn’t survive contact with production, I’m not interested in pretending it worked.
- Systems, not demos. I care less about what a model can do in one impressive session and more about what an organization can repeatedly trust it to do. The demo is usually the easy part.
- Organizational transformation. The interesting question isn’t whether AI can perform a task once. It’s how teams, roles, workflows, and responsibilities should change when AI can perform it repeatedly.
- The uncomfortable calls. I’ll say the thing most AI coverage dodges. Some of it will annoy engineers, some of it will annoy managers, some of it will annoy people selling transformation from a slide deck. That’s usually a sign it needed saying.
- Written by a human. No AI filler. No “in today’s rapidly evolving landscape.” Just how I’d explain it to another engineering leader over coffee.
Who this is for#
Engineering leaders responsible for real delivery. Senior engineers adapting to AI-driven execution. Platform and infrastructure leaders. Founders building technical orgs. Anyone moving AI from demos into production and deciding how humans and agents should work together.
If you’re running a real org, operating real systems, or making real budget and architecture decisions in the AI era, we’ll get along.
If you’re looking for prompt tips, generic AI enthusiasm, or feel-good transformation content, this isn’t the place. That’s not a shot at anyone. It’s just not what I do.