Code was never the point.
I’ve led 150 people and I’ve led none.
The real work has always been closing the distance between what someone means and what a system actually does.
AI compressed almost every part of building software. It did not compress that distance. It made it the only thing that matters.
That gap is where I’ve spent thirty years.
Most executives at my level made that gap somebody else’s job.
I’ve been a CTO four times. I’ve also written production code every one of those years — including this one.
Not as a hobby. Because the leaders who can still feel the material make better calls about it.
AmazonTote, 2010: pilot from zero to launch in 10 weeks — interface, logistics, operations, and global procurement of materials. One designer, two engineers, no AI. Small and fast was always the method.
- 47years programming — started at 5, on a Commodore
- 3MS degrees — Computational Engineering, Software Engineering, Technical Japanese
- 5patents (Amazon, last-mile delivery)
- →Shipped code this week
Boeing 787 in 2005: I introduced test-driven development and continuous integration to commercial aerospace. That’s the pattern — bring the discipline in before it’s fashionable.
Or I right the ship — and build the team that keeps it upright.
I’ve never been only an executive or only an engineer. The two have run in parallel the whole way. This is the other track: distributed organizations, up to 150 people, five countries, remote-first since 2015 — before it was mandatory and after it stopped being fashionable.
- —Rebuilt product and engineering orgs from the ground up at both Varsity Tutors and RealSelf, including VP and Director hires out of Amazon
- —Authored the work-from-home and remote-first policies; ran the transition
- —Turned QA from an isolated role into a core engineering discipline
- —Introduced leveling, banding, and quarterly calibration so people could see their own path
- —10+ engineers have followed me between companies. Some for the third and fourth time. Three seasoned principal engineers followed me to Lium.
The last one is the number I’m proudest of, and the only one on this deck I didn’t have to work to produce. People vote with their feet.
Scale the team. Break the monolith. Every company, same move.
This is not a highlight reel. It is what I do when I arrive.
| Sears Home Services | Varsity Tutors | RealSelf | Solo practice | |
|---|---|---|---|---|
| Arrived to | No engineering org, no digital product | Monolith. 32% of delivery online. Pre-Series C. | 100+ person org, margin pressure, flat traffic | Clients with no team and a deadline |
| Team move | 0 → 30 engineers in 10 months | Rebuilt org to 50 distributed; banding for IC and management; QA into core engineering | Rebuilt product and engineering org; banding and company-wide calibration; remote-first transition | 1 operator + agents |
| Technical move | TDD/BDD, multiple daily deploys | Monolith → SOA; CI/CD; 60+ services; 16 AWS accounts | Monolith → SOA; React and micro-frontends; analytics modernization | Agentic SDLC end-to-end |
| Result | Two products launched in 10 months. 15% and 5% MoM growth. | Delivery 32% → 70% online. $2M annualized cost out. Revenue 3×. IPO at $1.4B EV. | +3900 bps operating profitability. Traffic 2× YoY. Lighthouse 30 → mid-80s. Defect detection: months → 15 minutes. | Production client systems, single operator |
Varsity and RealSelf are the same engagement run twice: rebuild the org, introduce banding and real career structure, level up engineering leadership, break the monolith into services. Two different industries, two different decades, same result. The fourth column is what that method looks like now that the tools changed.
One operator. An agentic SDLC. Systems that used to take twelve people.
Since 2024 I’ve run a production practice as a single operator with AI agents — design, architecture, build, deploy, operate. Not prototypes. Client systems in production.
100% of the backend schema, APIs, admin web, and consumer mobile app. Two apps live on iOS and Android. Solo.
The agent harness — orchestration that lets LLMs reason across databases, documents, and live APIs in one query. Automatic source indexing. Compounding artifacts. Terabyte-scale compute provisioning.
Client systems delivered end-to-end since 2024. Production, not prototypes.
This is not “I use Copilot.” This is an operating model with a defect rate I can defend.
Find the truth myself. Write something that runs. Hand it over on purpose.
- ✓I go to the source — the data, the code, the customer — before I have an opinion
- ✓Working software over decks. A running thing beats a convincing argument.
- ✓I build teams that outlast me. Handoff is designed in from day one.
- ✓Psychological safety as an engineering control, not an HR poster: people who feel safe report defects early, and early defects are cheap.
- ×AI as a press release
- ×Reorgs as a substitute for strategy
- ×Remote-in-name-only
- ×Heroics as a business model
I’d rather lose the engagement than take one where the answer is already decided.
Three shapes. Pick the one that matches your problem.
You have: several companies, uneven engineering quality, no time to fix each one.
I do: diagnose fast, level up leadership, SDLC, CI/CD, delivery cadence. Across the portfolio or embedded in one.
You have: revenue, a team, and a ceiling you can’t get through.
I do: org and technology development. Leader of leaders. The engine two slides back.
You have: an idea, a market, and no time.
I do: small and mighty. I build it. You sell it.
Terms — Remote-first. Available for one week per month in Seattle, Portland, or the Bay Area. Broader on-site possible from 2027.
Stages — Pre-revenue through Series B+. PE and VC portfolio work. Enterprise for the right build.
Start with a conversation.
I’ll tell you within an hour whether I’m the right person — including when I’m not.
wardvuillemot.com