Andrés Fuentes · Engineering Leadership
Twenty-plus years building engineering orgs and the platforms under them. I helped grow Zynga Austin’s QA center from 10 to 70, built the event-driven platform still running Blackbaud’s Giving Tuesdays, and cut about $8M a year in operating costs across Blackbaud and H-E-B. The teams did that work. I set direction, cleared blockers, and grew the people doing it into senior and staff engineers. Now: independent engineering leadership, including fractional VP of Engineering for an AI startup.
I started where software gets broken on purpose: testing games at Microsoft Game Studios through Volt, including the Xbox 360 launch team. QA is the best education in failure there is. You learn how systems break, and how people behave when they do.
Twenty years later that’s still the job description. The titles changed; the question never did: where does this break?
Day 1Xbox 360 launch team · 2005
Zynga in the FarmVille and CityVille years: a fifth of all Facebook page views, 50 million people playing every day, millions of dollars a day in revenue, an IPO mid-flight. I helped grow the Austin QA center from ten to about seventy, six teams at peak, and quality for the titles making that money ran through my teams.
We A/B tested everything, with mathematicians, psychologists, and data scientists arguing over which shade of blue earned the click. But what I fell in love with was growing the people and the org itself. That’s where leading stopped being a title and became the work.
20%of Facebook’s page views · 2011
Chaotic Moon was Austin’s wildest design-and-dev studio, the one Accenture would buy in 2015, and it had no QA org. I built one from zero: twelve people testing for Samsung, Univision, John Deere, Disney, and Marvel.
Then we turned quality into product: QA sold as a service, Jira and Confluence stood up for clients, mobile CI/CD pipelines back when almost nobody had one. New revenue, not just fewer bugs.
12person QA org · sold as a service
Blackbaud was the pivotal one. The company was breaking monoliths like Luminate Online and Raiser’s Edge NXT into event-driven services, with Confluent and ThoughtWorks at the table.
I led six teams, 50 engineers across Austin and Charleston, owning the whole giving path: donation forms, one-time and recurring gifts, receipting, and the Kafka layer underneath, with hundreds of millions of messages a day and a petabyte ingested before Kafka hit v1.0. The recurring installment engine we launched in 2019 has since scheduled $1.7B across 25 million payments on the same architecture, and every Giving Tuesday the platform helps move $60M-plus in a single day, war rooms and all.
Add an AWS-to-Azure migration, TDD as a way of life, a custom deployment platform, and $4.2M a year cut from cloud spend, which also turned the eight engineers who ran our own Kafka cluster into a product team. It’s the job that taught me the most, with the best people I’ve worked with.
$60M+moved in one Giving Tuesday
I left Blackbaud to prove what I’d built wasn’t a one-company trick, the same reason I’d once left Zynga for Chaotic Moon. At Dealerware I ran Application Engineering across frontend, backend, and native mobile: three architects, a manager, three scrum teams, a new career ladder, and a recruiting pipeline cut from four weeks to a week and a half.
At Shippo I led the eight-engineer platform team owning reconciliation, the last stop in the shipping pipeline, catching mistakes and price discrepancies at scale. Different domains, same result: calmer teams, faster delivery. Across Blackbaud, Dealerware, and Shippo I carried PCI-DSS, SOC 2, and GDPR compliance, audits included, plus HIPAA technical safeguards for Blackbaud’s healthcare-adjacent nonprofit clients.
4w→1.5wrecruiting turnaround · Dealerware
Texans take H-E-B personally, and its engineering carries the same stakes. I ran Event Order Management: the platform where business development managers stand up manufacturer deals, coupons, and order logistics; services responsible for $4 billion a year. My teams cut $3.8M a year in operating costs and $2.4M a year in storage while defect turnaround fell from fifteen days to under one.
The fix wasn’t heroics: severity-tiered OLAs, standing rituals with QA and Customer Success, and a critical response process where none had existed. Speed is a system property, orgs included.
15d→<1ddefect turnaround
Since late 2024 I’ve run my own engineering leadership practice for teams modernizing production systems. At AI Grants, an early-stage AI startup, I’m the fractional VP of Engineering: technology direction, org structure, hiring standards, and architecture review while those decisions are still cheap to change. At Umbrella Alliance US (Umbrella ND), a neurodivergent-led nonprofit, I’m the technical lead: I shipped a rebuilt web presence and AI tooling their two-person staff maintains on their own.
I’m in these tools every day, so other engineering leaders ask me what actually works. The answer is usually unglamorous: context files too big to help, a review step that shares the author’s blind spots, an agent with no test gate in front of it. So I build the fixes in the open: Dope TDD, a Claude Code plugin that won’t let code in until a failing test exists, and Agent OS, a team of AI agents with clear roles and a reviewer that reads the work cold.
Dailyin the tools I build
A template that runs a team of AI coding agents with clear roles, checked procedures and CI gates.
Updated
Agent OS on GitHubA Claude Code plugin that makes AI agents write the failing test first.
Updated
Dope TDD on GitHubNeed an engineering leader who still ships?
Open to Engineering Manager, Director, and VP of Software Engineering roles.
contact.afuentes@gmail.com