Fractional software engineer
I build the machine that ships your software.
I take over platforms that have eroded to the point where no one can safely change them, and turn them into a delivery operation that runs on one meeting a month.
Release cadence
between production releases
Client-reported issues
per month
Deploy time
push to production
Remote · GMT-3 · one human name in the commit log across 10 repositories since 2024.
Where the machine ran
You know the system is the bottleneck. You just can't prove it in a meeting.
Your last release was a quarter ago, and everyone still remembers it.
Nobody touches that one module without clearing the afternoon.
The engineer who understood the system left, and the documentation left with them.
Every deploy is an event, with a rollback plan and a prayer.
Features arrive three to six months after you approved them, if they arrive.
If you nodded twice, we should talk. If you nodded at all five, we should talk this week.
Three ways to work with me
Rescue
I reconstruct how your system actually works, then tell you what to fix and in what order.
- Ends in a written map: what's broken, what it costs you, what to do first.
- No rewrite proposal. A rewrite is what got the last team into this.
- You keep the map whether or not you hire me for what comes after.
Delivery Machine
Requirements go in, reliable releases come out — and the technical overhead stops being yours.
- Pipelines, test suites, staging-to-production promotion, monitoring, release notes on every deploy.
- Modernization in increments, alongside feature delivery. No freeze.
- One meeting a month. Everything else is written and async.
Build
From requirement to production, one person, no handoffs.
- Discovery with your team, then database, backend, frontend, and deployment.
- The delivery machine is installed from day one, before the erosion starts.
- No perfectly-written ticket required before work begins.
What it looked like in production
Medical education platform — medical board certification exam prep
Fractional engineer · sole owner of ten repositories, three in active development
Arrived
Ten repositories with no automated pipeline anywhere. A release roughly once a quarter, delivered inconsistently and with heavy rework. An approved request took three to six months to reach production. Two people on support hand-sent hundreds of personalized emails every week.
Built
The full delivery machine — automated tests, CI/CD with staging gates, contract validation, monitoring, and release notes pushed on every deploy — plus an AI-native development workflow running under my sole architecture and review, and a modernization of the codebase carried out in increments alongside feature delivery.
Result
Deploys went from two hours to fifteen minutes. The monthly follow-up operation went from a month of manual work to about two hours. Client involvement: one meeting a month.
Newspaper platform modernization — via Muvz
Senior engineer and architect · team of eight developers
Arrived
A legacy Java EJB monolith mid-modernization, already three months behind schedule.
Built
Five Spring Boot microservices extracted incrementally — no rewrite, no freeze — with event-driven integration over Apache Kafka, a centralized back-office for shared configuration, and engineering practice (Hexagonal Architecture, SOLID, DDD) established across the team.
Result
The three-month delay eliminated and delivery back on schedule, with the practice left behind in the team.
Presidential campaign platform — via Conddiz
Senior engineer and frontend architect
Arrived
A national campaign with a fixed, immovable date and no platform.
Built
One backend serving three frontends — the official website plus two production PWAs — with integrations across every major social network.
Result
Sustained the peaks at the most critical moments of the calendar, with no date negotiation.
Numbers marked ● are measured from git, CI and the test suite. The rest are counted or recalled from the state before the work.
Your involvement is one meeting a month.
Requirements, architecture decisions, and backlog live in the repository, as one source of truth you can read without me in the room. Every deploy pushes release notes — technical and business — to you automatically. You get proactive, high-signal updates instead of status calls, and you own the code and the pipeline from the first day, not from the last one.
Timezone
Async-first, remote, GMT-3 — overlapping business hours with the Americas and half of Europe.
Cadence
One scheduled meeting a month. Anything urgent has a channel; nothing urgent needs a calendar invite.
Decisions
Written down where the code is, so the next person — including future you — can reconstruct why.
Ownership
For 2 years, the commit log across all 10 repositories shows one human name: mine. The only other author is an AI agent I built the harness for.
Testimonials
from my most demanding clients — the ones who sit on the keyboard
“Miau miaaaaaau! Fsssssst!”
Lesada · Cat
“Miau.”
Bugada · Cat
“I have no idea how I ended up here.”
Leão Lascado · Taxidermied lion
Recent writing
Three places I write
BlogEssays on delivery, architecture, and one engineer running a platform with AI agents.Workaround-Oriented ProgrammingA living technical book on the gap between the architecture we present and the workarounds we ship. In Portuguese.Codex TechnomanticusThe working notes I keep for full-stack engineering. In Portuguese.Let's look at your system.
Thirty minutes, no slides. You describe what's breaking; I tell you what I'd look at first and whether I'm the right person for it.