Senior AWS work, delivered without drama.

I do cloud infrastructure work for companies where a bad afternoon in AWS becomes a bad week for the business. Most of my current engagements are multi-year: the industrial waste-equipment manufacturer I've been the sole AWS provider for since late 2024; the ISO-certified logistics platform I meet with twice a week; the professional-services team I've been working with since 2019 across two previous projects before the current one.

The short version of how I work: I make the architectural call, write down why, verify AI-generated reports against live data before acting on them, and keep durable runbooks so context survives tool and team changes. I bill conservatively relative to the value delivered — and I would rather send a client a documented "nothing to do here, but here's what I checked" than invent billable work. Clients keep calling, so the math works.

I calibrate communication to the stakeholder. A non-technical CEO gets the hand-holding version and a decision-ready recommendation. A risk-averse senior engineer gets methodical and educational, with citations. A no-nonsense manager gets the answer with no extraneous talk. All three relationships outlast any single project.

How I approach the work

• Risk-aware execution. Backups before upgrades. Dry runs. Dev-and-staging-first patterns. Rollback paths written down, not improvised in the moment.
• Cost discipline. Cloud bills are an engineering problem. I've saved clients recurring thousands per month through right-sizing, retention tuning, and architectural choices — and killed a $355/mo line item on my own product.
• Documentation rigor. Decision logs, infrastructure inventories, and runbooks so the next engineer — even if it's future me — can pick up context fast.
• Independent judgment. I'll make the executive call when one's needed, and own it. That's usually why clients keep me around.
• Verification habit. I independently check AI-generated reports against live data. I've caught factual errors in client-supplied analyses. The tool supports the analysis; it doesn't replace it.