Software developer and AI engineer, from data to production

---
title: About | Fernando Moreira Costa
url: https://nandomoreira.dev/en/about
---

# Software developer and AI engineer, from data to production

I have been building software end to end for 15+ years, from the interface to
the data, and today I ship AI agents and AI products to production. At
Liquid AI, I owned the mortgage data pipeline for more than 20 real estate
developers, where I cut the time to prepare a dataset and deliver it into
the pipeline from about five days down to as little as 30 minutes. I
sustained that operation with my own AI-assisted development framework,
with security guardrails and mandatory quality gates. Before that, six
years on high-scale products at Itaú Unibanco, Pipefy and Skyone, and open
source since 2012.

I build the product, the data pipeline and the engineering standard that
make AI deliver measured results in production, not only in demos.

## Results

- **5 days → 30 min** (to prepare a client dataset and deliver it into the pipeline): Liquid AI. From about five days down to as little as 30 minutes, replacing
case-by-case manual work with API-driven extraction and pre-normalization
assisted by AI.
- **20+** (real estate developers on the data pipeline I operated): Liquid AI. Mortgage datasets that become metrics inside the product.
- **0% → 92%** (coverage on pipelines that shipped to production with no tests): Liquid AI. 920 cases written across two pipelines, which reached 84% and
92%, with a coverage gate in CI and deploys blocked when the suite fails.
- **4.800+** (test cases protecting the pre-normalization engine): Liquid AI. Unit, integration, regression and parity tests, against silent
breakage on every data transformation change.
- **15+ years** (building software, from agency to bank): Since 2011, with six years on high-scale products at Itaú Unibanco,
Pipefy and Skyone.
- **2012** (the year I started publishing open source): 476 followers and 122 public repositories on GitHub, with Rust tools
published on crates.io.

## How I can help

### End-to-end product

For anyone who needs to take a product off the whiteboard without
staffing a whole team. Interface, backend, database and infrastructure,
from opportunity to production, with the test written before the code.

- Case: [BragStack](https://nandomoreira.dev/en/portfolio/bragstack)

### AI and data in production

For anyone with data arriving in different formats and manual process
blocking operations. Data pipelines and agents that automate the
repetitive work, with a regression suite protecting every change.

- Case: [Financial data normalization](https://nandomoreira.dev/en/portfolio/financial-normalization)

### AI engineering standard

For teams that already use coding agents, but with no standard or
control. An AI-assisted development flow with mandatory TDD, quality
gates and executable security guardrails.

- Case: [Amnesia](https://nandomoreira.dev/en/lab/amnesia-claude-code)

## Experience

### AI Engineer • Full Cycle Developer

- nov/2025 – sep/2026 · Liquid AI
- I owned the data pipeline that turns client real estate developers'
datasets into metrics inside Liquid's products, and the AI-assisted
development framework that sustained software delivery.

- Cut the interval between receiving a dataset and delivering it pre-normalized into the pipeline from about five days down to as little as 30 minutes, with API-driven extraction and pre-normalization assisted by AI.
- Designed and built Play Manager, an internal product that became the source of truth for the operation, with phase, owner, deadline and health of every dataset visible in one place for the team and leadership.
- Covered with tests two pipelines that shipped to production with no automated tests, 920 cases taking coverage from 0% to 84% and 92%, with a CI gate and deploys blocked when the suite fails.
- Built the internal AI-assisted development framework on Claude Code, with 106 automations, 34 commands, 19 specialized agents, mandatory TDD per tool and 21 inviolable security rules.
- Orchestrated three AI CLIs with separated roles, under the rule that whoever writes does not judge, and the orchestrator is the only one who validates against the gates and versions.

### Technology Manager

- aug/2025 – oct/2025 · Norte Marketing
- I led the technology area as the team's main technical reference,
owning architecture, technical design, documentation and delivery
management.

- Rolled out an adapted Scrum agile culture, integrating Jira, Bitbucket and Confluence to make deliveries predictable.
- Automated versioning and deploys in AWS pipelines, with linters and automated tests as a quality gate.
- Implemented preview environments per Pull Request, shortening the QA validation cycle.

### Senior Frontend Developer

- dec/2024 – apr/2025 · Skyone Solutions
- React applications for digital products used daily by thousands of
users.

- Built a design system and reusable component library, cutting rework across squads.
- Optimized performance with a focus on Core Web Vitals, technical SEO and accessibility.
- Led code review, defined development standards and mentored junior and mid-level developers.

### Senior Software Engineer

- apr/2022 – jun/2024 · Itaú Unibanco
- Microfrontend products for the bank's unauthenticated pages, at the
intersection of Marketing and Technology.

- Implemented personalization and A/B testing with VWO and Salesforce Interaction Studio for the unauthenticated audience.
- Raised accessibility, performance and clarity of the public journeys, in an environment of very high traffic and strong regulatory demand.
- Translated marketing demand into sustainable frontend architecture, alongside the business teams.

### Senior Fullstack Developer

- jun/2020 – nov/2021 · Risknow.ai
- Data intelligence dashboards applied to the mortgage credit journey,
integrating APIs, microservices and WhatsApp bots that shortened the
property appraisal and financing cycle.


### Senior Frontend Developer

- mar/2019 – feb/2020 · MedPrev e Trackmob
- Reactive interfaces in Vue.js and Nuxt.js consuming REST and GraphQL at
MedPrev. At Trackmob, Vue.js dashboards on Ruby on Rails for NGOs to
track campaign impact in real time.


### Senior Web Developer

- oct/2017 – mar/2019 · Nossa Causa e Workshop Agência & Escola
- Responsive and accessible WordPress sites for institutional clients and
NGOs, also acting as UX and analyst in planning the solutions.


### Senior Frontend Developer and UX Designer

- sep/2016 – oct/2017 · LojasKD Móveis
- Symfony e-commerce consuming microservices in Node.js, Python and PHP,
with user research and A/B testing after the promotion to UX Designer.


### Mid-level Front-end Developer

- jan/2015 – aug/2016 · Pipefy e MalaPronta.com
- I started the new interface of Pipefy's process management platform. At
MalaPronta.com, I modernized a five-year-old layout for responsive and
cut server load by cleaning and minifying CSS and JavaScript.


### Web Developer

- nov/2010 – may/2014 · Agência WX, IBPT, Salsa:Mobion, Totalize e Agência CGDW
- Career start in PHP, MySQL, WordPress and Magento: institutional sites,
online stores, proprietary CMSs and a drag-and-drop page builder for
WordPress.


## Education

- jan/2015 – dec/2016: Technologist in Internet Systems, Universidade Positivo
- jan/2010 – dec/2011: Technical in Computing, Colégio Estadual do Paraná

## Stack

- TypeScript
- JavaScript
- Python
- Rust
- Go
- PHP
- Ruby
- React
- Next.js
- Vue.js
- Nuxt
- Angular
- Astro
- Tailwind CSS
- Node.js
- Bun
- Express
- Hono
- FastAPI
- Laravel
- Symfony
- Ruby on Rails
- GraphQL
- PostgreSQL
- MySQL
- DynamoDB
- Firebase
- Polars
- AWS
- Google Cloud
- BigQuery
- Docker
- Playwright
- Vitest
- Electron
- WordPress
- Claude Code
- Codex
- Gemini CLI
- Grok
- Antigravity
- Cursor
- MCP

## About this site

This site is also a case study: it was built to be read by people and by AI agents, with the same content in both layers.

- **Built with:** Astro 7 rendered on the server, TypeScript and Tailwind CSS 4, published on Firebase App Hosting.
- **For agents:** An MCP server, AGENTS.md, llms.txt and a markdown version of every page, opened by the AI button in the corner of the screen. [How to connect](https://nandomoreira.dev/en/mcp)
- **Light by default:** A single browser script, under 1.5 KB, for the pointer effect and the copy buttons. Project art is SVG generated on the server.
- **Privacy:** No cookies and no analytics. The site only keeps what you send through the form. [Privacy policy](https://nandomoreira.dev/en/privacy)
- **How it was made:** Built with Claude Code, with tests written before the code and automatic gates for contrast, keyboard, 320px screens and data leaks.