The full journey

A finance job, businesses of my own and a sailing school, then the move into software, which began with studying. Each stop below left me something I still use.

Filter the map

Years

Tools

Concepts

Skills

2006 – 2022

Prologue

2006
2006 – 2012
SEBRAE and São João del-Rei
Business administration student, then company founder
Business came first: at SEBRAE’s technical school, then in a company of my own.
Concepts
Business administration
Business planning
Company formation
Skills
Entrepreneurship
Business management
Read the storyClose the story
Everything began at SEBRAE-MG’s technical school in Belo Horizonte, where I earned a technical diploma in business administration while finishing high school. There, in 2007, I wrote my first business plan.
Next I moved to São João del-Rei to study psychology, and stayed four and a half years. A friend there who produced music knew about my business training and came to me with an idea: a music production company, with him making the music and me running the business. Exploring that idea, in 2012, I registered my first company.
When I left São João del-Rei, I took the company with me. The same registration stayed with me through the ventures that followed, and today that company is HC Consultorias, my software company.
2013
Financial Assistant
2013 – 2017
Immersus Language School
Financial Assistant → Financial Coordinator
Back in Belo Horizonte, I joined the finance office of a language school and went on to coordinate it.
- Worked with the school’s director, a financial consultant and a marketing consultant on a team that more than doubled the school’s revenue.
- As financial coordinator, led the financial planning with the school’s management.
- Automated the management (BI) reports with Excel VBA.
Tools
Excel
VBA
Concepts
Financial planning
Management reporting
Business intelligence
Report automation
Strategic planning
Skills
Team building
Multidisciplinary teamwork
Data analysis
Stakeholder management
Read the storyClose the story
- Worked with the school’s director, a financial consultant and a marketing consultant on a team that more than doubled the school’s revenue.
- As financial coordinator, led the financial planning with the school’s management.
- Automated the management (BI) reports with Excel VBA.
When I came back from São João del-Rei, I joined Immersus as a financial assistant. The job was all numbers, and that is where analytical thinking and strategic targets stopped being theory for me.
At one point I left to try a business of my own, and I came back to take over the finances again. This time I grew into the coordinator’s role, planning with the school’s management and automating the management reports in Excel VBA.
The biggest change came from a group: the school’s director, an outside financial consultant, a marketing consultant and me, each seeing the same business from a different side. Together we more than doubled the school’s revenue. That group was my first multidisciplinary team: the first I helped put together, and the first I worked in.
Entrepreneur
Brownies, a bakery, a hostel
Entrepreneur
Three businesses, one after another: brownies on the street, a bakery, and a hostel near Brumadinho.
Concepts
Cost and pricing
Direct sales
B2B supply
Hospitality
Skills
Entrepreneurship
Sales
Customer relations
Risk-taking
Read the storyClose the story
The office wasn’t enough, so I left it. With a friend, I started baking brownies in batches and selling them on the street in person, one customer at a time, and every sale put a real number on cost, price and whether someone would come back.
From brownies I moved on to a bakery that made cakes, bonbons and sweets, supplying fairs and the breakfast tables of local guesthouses. Then I returned to finance at Immersus.
Even back behind a desk, I kept looking for the next venture. Living near Brumadinho, I saw the tourist traffic in the region and opened a hostel, run alongside the finance job with a partner in charge of operations. After I closed the hostel, sailing came along, and for a short while I kept both the boats and the finance job.
Sailing Instructor
Until 2020
Sailing, from Lagoa dos Ingleses to Lagoa Santa
Instructor, then partner, then school owner
I spent years on the water teaching people of every age to sail, ran a school’s operations and, in the end, opened my own.
Tools
Laser
Snipe
Tahiti 16
Optimist
Stand-up paddle
Concepts
Pedagogy
Teaching methods
Operations
Financial management
Institutional relations
Skills
Teaching
Active listening
Knowledge transfer
Leadership
Stakeholder management
English
Entrepreneurship
Read the storyClose the story
Sailing found me through English. A friend who had grown up around boats sold a beginner’s course to two British children, though he couldn’t speak a word of English, and called me in to translate while he taught. By the end of that course I was hooked, and soon I was working for him.
Teaching became the heart of what I did. I taught every age, from children to adults, and the job was less about the boat than about the person on board: listening first, finding the method that worked for each student and passing on, step by step, what I knew. Along the way I took the Brazilian Sailing Confederation (CBVela) courses. I taught Laser, Snipe, Tahiti 16 and Optimist classes at Nauttec, at Uaiwind and at the Clube de Regatas Afonso Ligório (CRAL), inside Minas Tênis Clube.
With time I became my friend’s partner at the school on Lagoa dos Ingleses. He brought the name and the know-how; I took on operations, the Laser classes and, later, the school’s finances. Sailing also meant dealing with its institutions (the Minas Gerais federation, FMVela; the Brazilian Sailing Confederation, CBVela; and the Brazilian Olympic Committee, COB), and moving anything forward among them was, above all, a matter of politics.
When the club at Lagoa dos Ingleses closed, I moved to Lagoa Santa and opened my own school, SUP Lagoa Santa, with boat storage, stand-up paddle rentals and sailing lessons. The school ran until 2020, when the pandemic stopped everything on the water.
2020
Uber Driver
2020 – 2022
Uber, Belo Horizonte
The switch
Driving for Uber paid the bills while I studied to become a developer.
Skills
Resilience
Tenacity
Customer service
Attention to detail
Read the storyClose the story
When the pandemic closed the school, I started driving for Uber in Belo Horizonte, and before long I was studying at the same time. Those years demanded a great deal of resilience and tenacity, and I still did the job with all the care I had, whatever the job: passengers rated me highly, and one review became a nickname, “the legend of the wheel”.
Studying slowly took over my hours, and by 2022 I had stopped driving, before finishing Trybe.

2021 – today

The story

2021
FullStack Dev
2021 – 2023
CTFL, Trybe and CS50
Software testing and full-stack web development
Software testing first, then full-stack development, then the computer science underneath it all.
- Prepared for the ISTQB Certified Tester Foundation Level (CTFL v4): test design techniques, static testing, test management and reporting, and testing in agile teams.
- Completed Trybe’s full-stack web development program: 1,500 hours, taught through projects.
- Took Harvard’s CS50, Introduction to Computer Science, to learn what happens below the code, starting with C.
- As a crowdtester on uTest and Crowdtest, found bugs in live products for Brazilian and international clients.
Tools
JavaScript
TypeScript
Node.js
Express
React
Redux
React Testing Library
HTML
CSS
MySQL
MongoDB
Docker
Git
Unix shell
C
Python
R
React Native
SQL
Concepts
Computer science
Algorithms
Data structures
Software testing
ISTQB CTFL
Test design techniques
Static testing
Test management
Agile testing
Crowdtesting
Exploratory testing
Low-level programming
ORM
Skills
Full-stack development
Test design
Bug reporting
Bug hunting
Self-directed learning
Test automation
Data analysis
Read the storyClose the story
- Prepared for the ISTQB Certified Tester Foundation Level (CTFL v4): test design techniques, static testing, test management and reporting, and testing in agile teams.
- Completed Trybe’s full-stack web development program: 1,500 hours, taught through projects.
- Took Harvard’s CS50, Introduction to Computer Science, to learn what happens below the code, starting with C.
- As a crowdtester on uTest and Crowdtest, found bugs in live products for Brazilian and international clients.
Testing came first. Preparing for the CTFL gave me far more than a way to find bugs: design techniques such as equivalence partitioning, boundary values, decision tables and state transitions, static reviews, test management, how to write a report someone can act on, and where testing fits in an agile team. Crowdtesting on uTest and Crowdtest then put all of it to work on real products.
Next I joined Trybe’s full-stack web development program, taught through projects. The program started with web fundamentals, Unix, the shell and Git, moved to the front end with React, Redux and React Testing Library, then to the back end with Node.js, Express, MySQL, MongoDB, ORMs and Docker, and closed with algorithms and data structures. That is where my main stack took shape, the one I still use most: Node.js, Express and React in TypeScript, with SQL.
Trybe was turning me into a developer, but I missed knowing what happens underneath, so in 2022, alongside the program, I took CS50, Harvard’s introduction to computer science. CS50 starts in C, with memory and how a computer actually runs a program, before moving on to Python and SQL. I don’t write C for a living, and that was never the point.
I also studied Python, R for data analysis and React Native, each to a different depth. Laravel came later, on the job.
2023
QA Analyst
Jun 2023 – present
Brickup
Software Tester, building the QA function
I joined a construction-tech startup as a manual tester and set up the QA function it didn’t have yet.
- Together with engineering, cut open defects from about 300 to under 100, ending the recurring talk of stopping everything for a bug smash.
- Incident and failure management: a defined path from bug report to engineering, error reporting and feedback inside the platform, and proactive alerts on API and interface failures.
- Wrote the QA documentation: bug-report standards, the protocol for opening tickets with engineering and the testing process itself.
- Created the end-to-end test project: Cypress running scenarios written in Gherkin.
Tools
Jira
Cypress
Cucumber
Gherkin
TypeScript
Docker
Docker Compose
MySQL
GrowthBook
OpenFeature
Datadog
ESLint
Concepts
Incident management
Bug report standards
Test documentation
Feature flags
Real-user monitoring
Proactive alerting
End-to-end testing
BDD
Unit testing
Integration testing
Regression testing
Exploratory testing
Black-box testing
White-box testing
Load testing
Static testing
Test prioritization
Release gating
In-app feedback
Skills
Quality strategy
Process design
Technical writing
Test automation
Team training
Read the storyClose the story
- Together with engineering, cut open defects from about 300 to under 100, ending the recurring talk of stopping everything for a bug smash.
- Incident and failure management: a defined path from bug report to engineering, error reporting and feedback inside the platform, and proactive alerts on API and interface failures.
- Wrote the QA documentation: bug-report standards, the protocol for opening tickets with engineering and the testing process itself.
- Created the end-to-end test project: Cypress running scenarios written in Gherkin.
Brickup was my first job in software, in June 2023: a construction-tech startup with a growing platform and no QA function yet. I came in as a manual tester, and the cost of having no QA showed in the backlog, where every couple of weeks someone suggested we stop everything for a bug smash.
Instead of just testing harder, I started with the process. I defined how a bug is reported and how it reaches engineering, wrote the standards and the protocol for opening tickets, documented how we test, and gave customers a way to report errors and send feedback without leaving the platform. With monitoring and proactive alerts on API and interface failures, we started hearing about problems before customers had to report them.
The techniques followed, applied where they paid off rather than by the book: white-box and black-box testing, static review, and unit, integration, regression, exploratory and load tests, all drawn from the CTFL syllabus and chosen by impact. Feature flags made releases safer, and I created the end-to-end test project, with its scenarios written in plain language, BDD style.
Along the way I trained the team in the new practices, and together we brought the backlog down from about 300 open defects to fewer than 100.
2023
Within my first year
Brickup
Scrum master, product owner and team lead
Taking on problems that had no owner yet kept turning into new roles.
- Rebuilt the Jira board and created the process for modeling and designing user stories.
- Wrote the specs and blueprints for the billing and financial services: Open Finance bank reconciliation, invoice issuance and ERP integration.
- Took over the public API documentation.
- Scouted talent, designed the hiring processes and, for a time, coordinated a team of five developers.
- Designed and deployed microservices, among them the integrations service that connects the platform to its CRM, billing and monitoring.
Tools
Jira
Confluence
Python
FastAPI
AWS Lambda
Serverless Framework
API Gateway
Docker
Docker Compose
HubSpot
Asaas
Datadog
Concepts
Scrum
Story modeling
Product ownership
Requirements engineering
Technical specification
API documentation
Microservices
Serverless
Billing
Bank reconciliation (Open Finance)
Invoice issuance (NFS-e)
ERP integration
DevOps
Feature modeling
Skills
Leadership
Product management
Talent scouting
Hiring
Team coordination
Architecture
Technical writing
Read the storyClose the story
- Rebuilt the Jira board and created the process for modeling and designing user stories.
- Wrote the specs and blueprints for the billing and financial services: Open Finance bank reconciliation, invoice issuance and ERP integration.
- Took over the public API documentation.
- Scouted talent, designed the hiring processes and, for a time, coordinated a team of five developers.
- Designed and deployed microservices, among them the integrations service that connects the platform to its CRM, billing and monitoring.
Once the QA function was running, new work came before any new title: first I was offered the scrum master’s role. I rebuilt the Jira board and created a process for modeling and designing user stories, so that what reached development was already clear.
From there the work moved upstream, into what a product manager or product owner would normally own: requirements, specs, blueprints and documentation. That is how I came to write the specs for the billing and financial services, with bank reconciliation through Open Finance, invoice issuance and ERP integration, and to take over the public API documentation.
Brickup was also where I started learning architecture and DevOps for real. My first commit to the main server containerized the local environment, and I went on to design and deploy serverless services, such as the integrations service that connects the platform to its CRM, billing and monitoring. Later I also helped design the API gateway.
Hiring came to me as well: I went looking for people and designed the selection processes, and for a while I coordinated a team of five developers.
2024
2024
Brickup
The AI era: Constructo and daily-report insights
Once AI stopped being a differentiator and turned into a requirement for success, programming became my most valuable skill. The first product was Constructo, a construction planner.
- Constructo saved two Black Friday campaigns and put innovation back on the startup’s agenda.
- Refactored Constructo with construction-planning techniques and usage analytics, taking it from a long beta to a product sold as part of the platform’s plans.
- Designed the AI that reads the history of daily construction reports, and worked closely with the developer who built it.
Tools
n8n
Google Gemini
JSON Schema
Concepts
Structured output
One-shot prompting
Construction planning
Product analytics
Data-driven decisions
Lock-in
AI product strategy
Workflow automation
Prompt engineering
Skills
AI product design
Product discovery
Product management
Read the storyClose the story
- Constructo saved two Black Friday campaigns and put innovation back on the startup’s agenda.
- Refactored Constructo with construction-planning techniques and usage analytics, taking it from a long beta to a product sold as part of the platform’s plans.
- Designed the AI that reads the history of daily construction reports, and worked closely with the developer who built it.
Constructo started as an automation in n8n: a form goes in, a one-shot call to Gemini returns structured JSON, and a finished construction schedule comes out.
Constructo saved two Black Friday campaigns and put innovation back on the startup’s agenda, then spent a year in beta. I refactored it, applying construction-planning techniques and basing every product decision on usage analytics, and the proof of concept became an MVP and then a product with its own place in the platform’s plans.
The second was the daily-report AI, which I designed and followed closely while another developer built it. The AI surfaces insights from the history of construction reports, spots patterns that hurt site operations and suggests improvements, and it builds lock-in by giving customers a reason to keep feeding the platform data.
2025
Tech Lead
Jan 2025 – Dec 2025
La Fabrique Flottante
QA → Tech Lead
Alongside Brickup, I joined a software house in a QA role and, by April, was leading its projects technically, first Beamble and then Dalegig.
- Worked directly with the client and bridged the international team and the Brazilian developers.
Tools
Jira
Figma
Excalidraw
Mermaid
Concepts
Solution design
Requirements engineering
Release validation
Wireframing
Delivery planning
Skills
Technical leadership
Client communication
Cross-cultural teams
English
Project management
Read the storyClose the story
- Worked directly with the client and bridged the international team and the Brazilian developers.
In January 2025, still at Brickup, I joined La Fabrique Flottante, a software house, in a QA role on Beamble, a live video shopping platform for high-end fashion built for a French client. By April I was leading the project technically, and the company later brought me into its other projects, such as Dalegig.
Leading there meant listening to the client, finding the problem behind each request, choosing techniques and tools, designing the solution, then setting deadlines and the development cadence, and seeing every delivery through. Most of that work ran through Jira, with low-fidelity mockups in Figma and Excalidraw and usage flows in Mermaid, and in English, between the international team and the Brazilian developers.
2025
Mar 2025 – present
HC Consultorias e Desenvolvimento de Software
Managing partner
The company I started in 2012 became a software company, with room to grow.
Concepts
Custom software engineering
Business scaling
Skills
Business management
Entrepreneurship
Client relations
Technical consulting
Read the storyClose the story
By 2025 my work had outgrown what an employment contract could hold, and I needed room to scale my own company. In March 2025, the company I had registered back in 2012 became HC Consultorias e Desenvolvimento de Software, a limited liability company with me as managing partner.
HC’s clients and partners have included Brickup, BID Tecnologia, La Fabrique Flottante and Marketing para Cartórios – AI, for custom software, technical consulting and technical leadership. HC is also where I build my own products.
2025
Solutions Architect
2025
Beamble
Solutions architecture and mobile v2
I took on most of the mobile v2 rewrite with the two mobile developers and moved the platform from Lightsail to a production-grade ECS setup.
- Mobile v2: rewritten by a task force with the two mobile developers, in which I took on most of the code.
- API: tangled code reorganized into a clean monolith, with media upload, messaging and scheduled jobs split into services of their own.
- Infrastructure: ECS behind a load balancer, VPC, RDS and blue-green deployments with CodeDeploy, all defined in CloudFormation, with the scheduled jobs on Lambda.
- Led the SEO and generative engine optimization (GEO) audit and its ten-part plan for visibility in search engines and AI assistants.
Tools
React Native
Redux Toolkit
VideoSDK
VisionCamera
Datadog
Laravel
Stripe
Firebase Functions
AWS Lambda
Serverless Framework
Amazon Lightsail
Amazon ECS
Elastic Load Balancing
VPC
Amazon RDS
CodeDeploy
CloudFormation
Next.js
Google Search Console
Bing Webmaster Tools
Python
Node.js
TypeScript
Concepts
Solutions architecture
Cloud migration
Auto scaling
Blue-green deployment
Infrastructure as code
Legacy modernization
Microservices
Event-driven architecture
Mobile performance
Real-user monitoring
Live video commerce
Payments integration
Multi-carrier logistics
Technical SEO
Generative engine optimization
Structured data
Core Web Vitals
Model Context Protocol
CI/CD
Proactive alerting
Server-side rendering
Skills
Architecture
Hands-on delivery
Performance tuning
Mobile development
Read the storyClose the story
- Mobile v2: rewritten by a task force with the two mobile developers, in which I took on most of the code.
- API: tangled code reorganized into a clean monolith, with media upload, messaging and scheduled jobs split into services of their own.
- Infrastructure: ECS behind a load balancer, VPC, RDS and blue-green deployments with CodeDeploy, all defined in CloudFormation, with the scheduled jobs on Lambda.
- Led the SEO and generative engine optimization (GEO) audit and its ten-part plan for visibility in search engines and AI assistants.
When I took over the technical side, Beamble was carrying a lot: live video shopping calls, a cart and checkout, Stripe payments, maps, and delivery through Uber, DHL and UPS in several countries. Underneath, the app had critical performance problems and outdated libraries, the API had grown tangled, and the platform ran on a single Lightsail server.
Then Google Play set a deadline: from November 2025, every update targeting Android 15 had to support 16 KB memory pages, which the app’s aging native stack couldn’t guarantee, and without that the app could stop shipping updates. That deadline was one more reason to get my hands into the code. In a task force with the two mobile developers, we rewrote mobile v2 in React Native. I wrote most of the code, tackling memory leaks, video performance and the maps, and added real-user monitoring with Datadog.
On the back end, the Laravel API went from tangled code to a clean monolith, with media upload, messaging and scheduled jobs moved into services of their own, on Firebase Functions and on AWS Lambda with the Serverless Framework. The platform left Lightsail for ECS behind a load balancer, inside its own VPC, with RDS, auto scaling, alerts and blue-green deployments through CodeDeploy, all defined in CloudFormation.
One of the pieces I’m proudest of from that year is the SEO and GEO audit: a study of how search engines and AI assistants see Beamble, turned into a ten-part plan, from structured data and server-side rendering to Core Web Vitals, tracking brand mentions in LLMs and a Beamble MCP server.
Every change had the same aim: a platform that scales, fails less, is easier for the team to work on and ships on time. Much of what made that possible I had first learned at Brickup.
2025
2025
La Fabrique Flottante · Dalegig
AI engine and interface lead
Brought into the software house’s other projects, I built the AI at the heart of Dalegig, a platform that connects live music with venues.
- Built the proof of concept for the musician interface, then coordinated and designed its third version.
- Built Chico, the AI agent that researches each musician on YouTube, Spotify, Instagram and the web and suggests goals to strengthen the profile.
Tools
React
Node.js
Nx
Supabase
PostgreSQL
Flowise
n8n
LangChain
SearXNG
Anthropic
OpenAI
Concepts
AI agents
Web scraping
Profile enrichment
Recommendations
Goal setting
Matchmaking
Proof of concept
Model Context Protocol
Conversational onboarding
Skills
AI engineering
Interface design
Team coordination
Full-stack development
Read the storyClose the story
- Built the proof of concept for the musician interface, then coordinated and designed its third version.
- Built Chico, the AI agent that researches each musician on YouTube, Spotify, Instagram and the web and suggests goals to strengthen the profile.
Beamble opened the door to La Fabrique’s other projects, and Dalegig came next. The platform connects musicians, artists and cultural producers with venues, and I was brought in for two things: an interface for musicians and an engine to help each profile stand out.
I started with the proof of concept of the musician interface and, by the third version, was coordinating and designing the interface. The engine is Chico, an agent that talks with each newly registered musician, researches them across YouTube, Spotify, Instagram and the wider web to build the profile, weighs reach and relevance, and turns what it finds into goals that appear next to the ones the musician sets.
Chico runs as agents on Flowise and n8n, with Anthropic and OpenAI models, inside an Nx monorepo with React, Node.js and Supabase.
2025
Jun 2025 – 2026
Prévia · Brickup and PHD Engenharia
Joint project lead
I led Brickup’s side of the joint project that grew into Prévia, an AI construction-planning platform built for large companies.
- A planning assistant grounded in Lean Construction, Advanced Work Packaging and PMBOK.
- A conversational agent that triggers and tracks the planning pipeline from chat: quantity takeoff, unit-price breakdowns, the work breakdown structure (WBS) and distribution.
- Custom n8n nodes that package LangChain’s deep-agents SDK, so agents with subagents can be built visually.
Tools
React
TypeScript
Redux Toolkit
Expo
n8n
LangChain
LangGraph
Zod
Google Gemini
Flowise
Supabase
AWS
React Native
Vite
Concepts
Lean Construction
Advanced Work Packaging
PMBOK
Work breakdown structure
Quantity takeoff
Unit-price breakdowns
Multi-agent systems
Human in the loop
Structured output
CI/CD
Quality gates
Construction planning
AI agents
Skills
Technical leadership
Domain modeling
Stakeholder management
Platform extension
AI engineering
Mobile development
Read the storyClose the story
- A planning assistant grounded in Lean Construction, Advanced Work Packaging and PMBOK.
- A conversational agent that triggers and tracks the planning pipeline from chat: quantity takeoff, unit-price breakdowns, the work breakdown structure (WBS) and distribution.
- Custom n8n nodes that package LangChain’s deep-agents SDK, so agents with subagents can be built visually.
Prévia was born as a joint project between Brickup and its partner and investor, PHD Engenharia, and I led it from the Brickup side, working with two of their engineers. The target was ambitious: construction planning for the companies behind the largest projects, from corporations and multinationals to mining, large real-estate developers and infrastructure works, on a platform ready to scale with them.
Planning at that scale calls for the methods large projects use (Lean Construction, Advanced Work Packaging and PMBOK), so the assistant applies them to each project’s parameters. The platform was designed to go well beyond housing, into industrial and port works, mining, energy and infrastructure.
Under the hood, the agent runs in n8n: a root agent with a methodology specialist as a subagent, partly on LangChain’s LangGraph, with Gemini models and human review where the numbers matter, such as extracting unit-price breakdowns. To make that kind of agent reusable, I packaged LangChain’s deep-agents SDK as custom n8n nodes. The platform around the agent is a React and TypeScript monorepo with an Expo mobile app, with CI/CD and quality gates on AWS.
2025
2025 – present
Immersus Language School
Technical Consultant
Back at Immersus, now as the school’s technical consultant.
- Coordinated the school’s three vendors on the technical side: the sales firm, the content marketing agency and the paid-ads manager.
- Reviewed the agency’s landing page for search visibility, design, responsiveness, performance and brand consistency.
- Built the school’s web platform: student, teaching and admin areas on their own domains, a content manager and two-factor authentication.
Tools
Laravel
React
TypeScript
Inertia.js
Vite
Tailwind CSS
shadcn/ui
Filament
Laravel Fortify
Sanctum
Spatie Permission
Concepts
SEO
Web performance
Server-side rendering
Multi-domain architecture
Role-based access control
Two-factor authentication
Delivery review
Responsive design
Brand consistency
Content management (CMS)
Skills
Technical consulting
Vendor management
Full-stack development
Read the storyClose the story
- Coordinated the school’s three vendors on the technical side: the sales firm, the content marketing agency and the paid-ads manager.
- Reviewed the agency’s landing page for search visibility, design, responsiveness, performance and brand consistency.
- Built the school’s web platform: student, teaching and admin areas on their own domains, a content manager and two-factor authentication.
In 2025 I returned to Immersus, this time on the technical side. The school had three vendors working on its growth (a sales firm, a content marketing agency and a paid-ads manager), and my role was to coordinate them and bring their work to a shared technical standard. The agency’s new landing page was the first test, and I reviewed it for search visibility, design, responsiveness, performance and brand consistency.
Alongside the consulting, I built the school’s web platform on Laravel and React through Inertia, with server-side rendering for the public site, Filament for content, and role-based access and two-factor authentication for the student, teaching and admin areas, each on its own domain.
2025
Nov 2025 – Jul 2026
BID Tecnologia
Technical Consultant
Technical consulting for a device trade-in company: a harness that lets its owner build software with AI, and a warehouse system in production.
- An AI development harness for a non-technical owner: rules, a mandatory flow from classifying the request to verifying the result, tests, coverage gates, hooks that block a commit and CI that blocks a merge.
- The warehouse system, the big win: inventory, triage, validation, disassembly, shipping, CRM and admin, in production.
- Consolidated a legacy spread across several apps into one monorepo.
Tools
Claude Code
Anthropic
React
TypeScript
Nx
Express
PostgreSQL
Supabase
Redis
Terraform
GCP
Vitest
ESLint
GitHub Actions
n8n
SearXNG
Firecrawl
Vite
Node.js
Concepts
AI development harness
Quality gates
Adaptive complexity
Architecture decision records
Row-level security
Multi-tenancy
Vision language models
Product roadmap
Go-to-market
Legacy modernization
Warehouse management
CI/CD
Infrastructure as code
Test coverage
Web scraping
AI agents
Skills
Technical consulting
Developer experience
Product strategy
AI engineering
Full-stack development
Read the storyClose the story
- An AI development harness for a non-technical owner: rules, a mandatory flow from classifying the request to verifying the result, tests, coverage gates, hooks that block a commit and CI that blocks a merge.
- The warehouse system, the big win: inventory, triage, validation, disassembly, shipping, CRM and admin, in production.
- Consolidated a legacy spread across several apps into one monorepo.
BID came to me through HC in November 2025, looking for help with its product roadmap and with how to take the product to market. BID’s trade-in and asset-management platform was spread across earlier versions and several separate apps, so the first job was to bring them into a single React and TypeScript monorepo, with row-level security for multi-tenancy and Terraform on GCP.
Over time a pattern emerged. The owner kept building prototypes in Lovable and bringing them to me to rebuild with security and scale in mind. Rather than redo every project, I set up a harness he could build in: rules the AI agent must follow, complexity levels that set how much process each change needs, mission logs that record every decision, and gates the code can’t skip, from tests and coverage to hooks that block a commit and CI that blocks a merge, all explained in plain language. He could keep building with AI, and what he built could reach production.
The big win was the warehouse system, which went into production. Alongside it, a pricing agent uses Claude’s vision capabilities to identify a device and its parts from photos, checks market prices and proposes an offer.
2026
AI Product Engineer
Jan 2026 – Sep 2026
Marketing para Cartórios – AI · D-Rocket group
Tech Lead and Full-Stack Engineer
An AI customer-service agent for Brazilian notary offices on WhatsApp, built with a team I coordinated, in which I owned the engineering from the first commit to the handover.
- Engineered the notary office panel, the admin panel and the agent itself, from architecture to production.
- Coordinated a multidisciplinary team, mentored a junior developer and took part in strategic decisions.
- Built for a regulated market, with every requirement of CNJ Provision 213/2026 mapped to the code and an eye on Brazil’s data protection law (LGPD) throughout.
- Handed over in September 2026 with the first two clients live, the system running, the team trained and operating procedures in place.
Tools
TypeScript
Node.js
Express
tRPC
React
Prisma
PostgreSQL
pgvector
Redis
n8n
OpenAI
Evolution API
Nx
Docker Swarm
Traefik
Amazon Lightsail
Amazon RDS
WhatsApp
Concepts
RAG
Vector search
Knowledge base curation
Speech-to-text
Multi-tenancy
Compliance
LGPD
Audit trail
Role-based access control
Two-factor authentication
Agent evaluation
Human in the loop
CI/CD
Product roadmap
AI agents
Usage-based billing
Skills
End-to-end ownership
AI engineering
Regulatory compliance
Team coordination
Mentoring
Product owner coaching
Technical leadership
Full-stack development
Knowledge transfer
Read the storyClose the story
- Engineered the notary office panel, the admin panel and the agent itself, from architecture to production.
- Coordinated a multidisciplinary team, mentored a junior developer and took part in strategic decisions.
- Built for a regulated market, with every requirement of CNJ Provision 213/2026 mapped to the code and an eye on Brazil’s data protection law (LGPD) throughout.
- Handed over in September 2026 with the first two clients live, the system running, the team trained and operating procedures in place.
Marketing para Cartórios wanted AI agents to serve Brazilian notary offices over WhatsApp, and in January 2026 I joined to build them. The product was a team effort: the founder, whom I supported as she stepped into the product owner role, domain experts from the notary world, and a junior developer I guided. Stack, token-based pricing and roadmap were defined with the founder.
Each notary office got a knowledge base built from its reference documents, such as laws, fee tables and opening hours, curated from the admin panel and searched with PostgreSQL and pgvector. Voice messages are transcribed before the agent answers. The agent is orchestrated in n8n, reaches WhatsApp through the Evolution API and writes its replies with OpenAI models. Before any change goes live, an evaluation tool scores the agent against scenarios and personas.
A regulated market asks for more than a chatbot. The platform was built around CNJ Provision 213/2026 and the LGPD, with mandatory two-factor authentication, one active session per user, an append-only audit trail, isolation between offices and an AI disclosure at the start of every conversation. Each requirement of the Provision is mapped to the code and tracked on the roadmap, and the agent evaluation includes personal-data scenarios.
2026
2026 – present
Brickup, today
Building BIA
Still at Brickup, now building BIA, a conversational financial assistant on WhatsApp.
Tools
n8n
WhatsApp
Evolution API
Concepts
Conversational AI
Tool calling
Agent routing
Multimodal input
Financial workflows
AI agents
Skills
Mentoring
Knowledge transfer
AI engineering
Read the storyClose the story
Today I support the QA engineer I trained to take my place, and much of the credit is his, because he is very good.
My own focus has moved to BIA, an agent in n8n with a router and tools to look up records, read the media people send and create draft financial entries, all from a WhatsApp conversation.
2026
2026 – present
My own products · HC Consultorias
AI Product Engineer
Still an entrepreneur: products of my own, from idea to code.
- A cloud backup for CCTV recorders, and a shared base for mini-games and mini-apps.
- WAI Framework: a six-phase method for building software with AI agents, from discovery to monitoring.
- Triadora, the newest, built to bring AI customer service within reach of any business.
Tools
TypeScript
Node.js
Express
tRPC
Prisma
PostgreSQL
React
Nx
Docker
AWS CDK
Amazon S3
ffmpeg
pg-boss
Claude Code
Concepts
Decision trees
Guardrails
Context management
Human in the loop
Prompt engineering
Provider-agnostic AI
Hexagonal architecture
Infrastructure as code
AI-first development
RAG
Model Context Protocol
AI development harness
Video processing
Background jobs
Skills
Product engineering
Entrepreneurship
AI engineering
Architecture
Read the storyClose the story
- A cloud backup for CCTV recorders, and a shared base for mini-games and mini-apps.
- WAI Framework: a six-phase method for building software with AI agents, from discovery to monitoring.
- Triadora, the newest, built to bring AI customer service within reach of any business.
With AI I still design, write and review the code. My focus is applications and features powered by AI, built with harnesses, context management, guardrails, retrieval and prompt engineering, where people make the decisions and the AI is a well-orchestrated tool in human-in-the-loop flows.
My first products were a cloud backup for CCTV recorders, built on a hexagonal architecture with ffmpeg, S3 and infrastructure in AWS CDK, and a shared base for mini-games and mini-apps. The method behind them became the WAI Framework, six phases from discovery to monitoring for building software with AI agents.
Triadora, the newest, follows the same principle of people deciding and AI assisting: the conversation path is an explicit, auditable decision tree, and the model only interprets what the customer says. The provider is a detail: one model can replace another, or none at all, without touching the flows.
2026
2026
This portfolio
Design, 3D and engineering
A 3D bust built from Blender recipes written in code, with nine lives in a single hero.
Tools
React
TypeScript
Three.js
React Three Fiber
Blender
glTF Transform
GSAP
Tailwind CSS
Vite
Cloudflare Workers
Cloudflare D1
Cloudflare Turnstile
Playwright
Vitest
ESLint
Knip
Claude Code
Python
Concepts
3D on the web
Code-driven 3D pipeline
Performance budgets
Content Security Policy
Edge computing
Static rendering
AI-first development
Accessibility
Skills
Creative development
Art direction
Agent orchestration
Test automation
Read the storyClose the story
The idea for a portfolio like this had been with me for years, and 3D always made me curious. I had played with Blender, but becoming a 3D modeler was never the goal, so the idea never went live.
What changed was the tooling. With today’s models and the WAI Framework, I could run the project like a small studio, with AI agents filling its roles: each life of the hero went through direction, modeling and review before my approval. Everything ships from a single Cloudflare Worker, with tests, performance budgets and a strict content security policy.
The hero is the creative side, and this page is the story behind it. The source code is public.
View source on GitHub

The full journey

SEBRAE and São João del-Rei

Immersus Language School

Brownies, a bakery, a hostel

Sailing, from Lagoa dos Ingleses to Lagoa Santa

Uber, Belo Horizonte

CTFL, Trybe and CS50

Brickup

Brickup

Brickup

La Fabrique Flottante

HC Consultorias e Desenvolvimento de Software

Beamble

La Fabrique Flottante · Dalegig

Prévia · Brickup and PHD Engenharia

Immersus Language School

BID Tecnologia

Marketing para Cartórios – AI · D-Rocket group

Brickup, today

My own products · HC Consultorias

This portfolio