Engineering and AI leader. 28 years in technology, 17 building enterprise platforms, 14 leading cloud transformations.
Building platforms, and the teams that run them.
2025-Present
Founder
Context Engineering for AI-assisted development, running in production at a Fortune 500 hospitality company. 16-hour tasks take 20 minutes.
2022-2024
Interim Director of Cloud, AWS ProServe
Stepped in to run the cloud org when the CIO asked for help, then led 55 engineers across 5 scrum teams. Account vending went from 30+ days to 45 minutes, and the CIO presented the work in an Innovation Talk at re:Invent 2023.
2019-2022
Sr. Director of Architecture
Promoted from Principal Architect within nine months. Led 8 architects plus the platform engineering and UX teams. The SEED platform took provisioning from 12-18 months to minutes.
2016-2017
Chief Architect of DevOps, Xentaurs
Ran engagements including Liberty Mutual, ASU, and Availity, and designed the firm's DevOps philosophy. Took over Liberty Mutual's struggling Docker migration and built the Fusion Platform, scaling from 3 applications to 300+ services with hundreds of daily deployments.
2014-2015
Director of Cloud
Built Utopia, a container platform with 100x better security defect density than the enterprise. The CISO made Docker a corporate mandate.
2012-2014
Manager of Cloud Systems Development
Promoted from architect to manager. Built and led the team behind Nibiru, a self-service AWS platform used by 70 development teams, with Gene Kim consulting on the transformation.
Building platforms, in public.
Designing HA infrastructure on MongoDB from 1.4 — failover bugs reported upstream, 20M active mobile sessions at 28,000 TPS proven in third-party testing.
Why there is no such thing as a "DevOps team" — and what managing high-performing engineers actually takes.
Nibiru, a self-service AWS platform used by 70 development teams — platform engineering before the term existed.
The argument that DevOps became the waste it was meant to eliminate — the thesis that led to everything since.
on outcomeops.ai ↗
How to measure an engineering organization by outcomes instead of activity — perfect pipelines still fail the business when you track the wrong metrics.
on outcomeops.ai ↗
Where the idea gets its name: the shift from DevOps practices to context engineering for AI-assisted development.
on outcomeops.ai ↗
Why buying AI coding licenses is not an AI strategy — repo-bound specs speed up one team, while enterprises need systemic acceleration.
Where the argument landed: context engineering for AI-assisted development, and the platform built to answer it.
I ran the same feature build on Grok twice. Between runs I changed nothing about the model or the prompt. I tightened one ADR and re-ingested it, and the second build fixed the governance violation on its own. Here is the before and after, including what the ADR could not fix.
Read moreYour agent didn't get dumber. Your process was never written down. Put decisions in ADRs and plans in files, and the model becomes swappable.
Read moreA month ago my AI music video pipeline made 30 seconds of one robot. Low Battery is 3:20, two characters, 42 directed shots, and five Shorts. Still no video editor. Here is the new process, shot by shot.
Read more