Breno Andrade
Contact

Software Engineer & Tech Lead

Breno Andrade

I lead an engineering squad at Brazil's largest workforce-management SaaS, working full-stack — backend, frontend, and deployment, end to end. On my own time, I've taken three products from idea to production, entirely by myself.

Experience

QuarkRH — Tech Lead

QuarkRH is Brazil's largest workforce time-tracking SaaS. I lead a squad responsible for core people-management modules:

  • Recruitment & Hiring
  • Performance Reviews
  • Climate Surveys
  • Training & Development
  • Internal Communication
Education

Academic Education

U

UNINASSAU

Postgraduate Lato Sensu Degree — Specialization in Computer Software Engineering

Jan 2026 – Dec 2026

U

UNINASSAU

Systems Analysis and Development

Jan 2024 – Dec 2025

Java, PostgreSQL, and 6 more skills

U

UFBA — Federal University of Bahia

Information Systems

May 2021 – Dec 2025

Python, Arduino, and 1 more skill

Stack

What I use day to day

Backend

  • Java
  • Spring Boot
  • PostgreSQL
  • RabbitMQ

Frontend

  • Angular
  • JSF

Infra & Deploy

  • Kubernetes
  • Grafana
  • CI/CD Pipelines

Engineering Practices

  • GitLab
  • Git Flow
  • Clean Code
  • Code Review

AI & Productivity

  • Prompt Engineering
  • AI-Assisted Development

Process

  • Jira
  • Confluence
Projects

Live in production

Catarina Fialho

Landing Page

A professional website for a clinical psychologist.

Visit site

Nabnak

SaaS

A study platform for students.

Visit site

Prontuário

SaaS

Medical records system for psychologists.

Visit site
Success Story

Nabnak — my biggest win

An AI study platform that turns any exam notice, syllabus, or PDF into an organized study plan in seconds — with a Kanban board, a built-in Pomodoro timer, and progress reports. I built it from scratch and it now has real, paying students.

653+

active students

4.96★

average rating

5,792+

AI-generated cards

19,000h

focus hours logged

Visit Nabnak

01 — The problem

Exam notices are a mess

Public exam notices and course syllabi show up as 40+ page PDFs — dense, repetitive, full of legalese. Students were losing hours just figuring out what to study before they could start studying.

02 — The approach

A pipeline, not a chatbot

I built a pipeline that breaks the PDF into topics, generates flashcards, and assembles a study plan with AI — then hands it off to a Kanban board with a built-in Pomodoro timer. A ready plan in minutes instead of a lost weekend.

03 — Trade-offs

Decisions I made on purpose

  • Async queue over a synchronous AI call

    Generating a full plan can take up to 30s. Instead of freezing the UI waiting on the model, I process it in the background and notify the student when it's ready — better UX, lower infra cost.

  • Cache by exam notice, not by user

    Public notices (INSS, TJ-SP, etc.) repeat across thousands of students. I cache the plan structure per notice and personalize only the progress layer — that cuts AI cost per user significantly.

04 — What's next

Observability over new features

With real paying students now, the priority is measuring the quality of generated cards — not just uptime — so a bad plan gets caught before a student complains.

I pasted the INSS exam notice and had a full study plan in two minutes. I used the Pomodoro timer every day and passed on my first attempt. Nabnak changed my study routine.

MC

Marcos C.

Passed the INSS exam, 2025

I used to rely on notebooks and kept losing track. With Nabnak I organized the entire TJ-SP syllabus into cards. The reports showed me exactly where I was falling behind. I passed!

RA

Renata A.

Passed the TJ-SP exam, 2025

I'm a medical student and used to rely on a generic note-taking app. With Nabnak, the AI builds my cards straight from the professor's syllabus. I save hours of planning every week.

FS

Felipe S.

Medical student, São Paulo

Featured

On LinkedIn

During my postgraduate studies, I built a JWT security module for Spring Boot. Once integrated, the controller simply receives the authenticated user through @UsuarioAtual — no manual token parsing, no repeated logic, no boilerplate.

23,550 impressions150 likes13 comments
Read the full post on LinkedIn
Contact

Let's talk

Open to talk about full-stack development, system architecture, or standardizing security across services.