Software & AI Engineering + International Relations · Brazil (UTC−3)

Software that ships, AI that holds up in production.

I’m , a Software & AI Engineer and International Relations undergraduate. I work on LLM-based systems, agent automation and QA for a US construction-procurement platform. I build the pipeline, the prompts and the tests that prove it all works — with an eye on the markets where software operates.

  • Software Engineering
  • AI Engineering
  • Prompt Engineering
  • QA & Automation
  • MCP
  • International Relations
Herickbuild · validate · shipSoftwareRails · APIs · CI/CDAI EngineeringLLMs · Agents · MCPQualityQA · AutomationGlobalInt’l Relations · ENfig. 01 — one engineer, four disciplines

02 / About

Engineering, AI and a global point of view, in the same person.

I work as a contractor at SubBase — a US construction-procurement SaaS platform — through Codengage. I own the human-in-the-loop data work, prompt engineering and QA for an LLM-based invoice-extraction pipeline, from the prompt to the test that validates it.

I believe good AI products are mostly good engineering: clear contracts, measured outputs, tests that catch regressions before customers do. Studying International Relations alongside Software Engineering is how I keep the other half in view: the people, markets and regulations the software has to serve.

Based in
Dois Vizinhos, Paraná, Brazil
Working at
SubBase (Fort Lauderdale, remote) · through Codengage
Studying
Software Engineering, UTFPR · International Relations, UNINTER
Languages
Portuguese (native) · English (professional) · Spanish, French (basic)
Available for
Talks, mentoring and tech communities

03 / Skills & expertise

Four pillars, one loop: build, validate, automate, ship.

  • Software Engineering

    Backend-leaning full-stack work, from data model to API to deploy.

    • Ruby on Rails
    • PostgreSQL
    • REST APIs
    • Docker
    • Git / GitHub
    • CI/CD
    • SQL / Redash
    • JavaScript
    • Astro & Tailwind CSS
    • Networking fundamentals
  • AI Engineering

    LLM pipelines that extract structured data, agents that act, and the prompt systems that keep them precise.

    • LLM pipelines
    • Prompt Engineering
    • AI Agents
    • MCP
    • Model integration
    • MLOps
    • LLM evals
    • Coding agents (Claude Code)
  • Quality & Automation

    Test plans, automated suites and system validation, treated as part of the product, not an afterthought.

    • Test strategy
    • Automated testing
    • System validation
    • HITL evaluation
    • Bug reporting
    • E2E testing in CI
    • Test management (MQ-Test)
  • Global collaboration

    Daily work in English with a U.S. team, plus International Relations training for the market side of software.

    • English (professional)
    • Remote-first
    • International Relations
    • Public speaking
    • Spanish & French (basic)
    • Mentoring

04 / Featured projects

Work that pairs AI with the discipline to prove it.

Featured · Open source

2025 — now

MQ-Test

A test-management platform (suites, cases, plans, runs, bugs, requirements) with an MCP server so coding agents can read and write test data. Integrates with GitHub, Linear and Slack; the automation repo stays the source of truth and MQ-Test becomes the dashboard that organizes and tracks it.

test cases organized
500+
runs tracked via MCP
9,000+
integrations
10+
  • Rails 8
  • PostgreSQL
  • Docker
  • MCP
  • GitHub API
  • Linear

Professional · SubBase

LLM invoice extraction, vendor by vendor

Vendor-specific prompt rules, a pre-extraction page-classification gate for mixed PDFs, schema migrations and structured bug reports for an extraction pipeline used in production.

  • Prompt systems
  • LLM eval
  • Redash / SQL

Automation

subbase-qa-tests

Automated end-to-end suite for the SubBase platform, organized in MQ-Test and run on CI, so regressions are caught before release.

  • E2E
  • CI

Product design

SUSAGENDAMENTO

Mobile app concept for scheduling public-health (SUS) appointments, designed in Figma.

  • Figma
  • Mobile UX

Agent workflows

prompt-pattern-docs

A skill that turns rough extraction problems into production-ready, documented LLM prompts.

  • Claude skills
  • Prompt patterns

05 / Journey

From cabling networks to shipping AI.

Work and study, in parallel, since 2023.

  1. Sep 2024 — now · Remote (Fort Lauderdale, USA)

    Data Entry / HITL / Prompt Engineering Specialist · SubBase through Codengage

    Contractor through Codengage, dedicated to the SubBase platform: the human-in-the-loop quality layer of an LLM invoice-extraction pipeline — vendor-specific prompt rules, behavior diffs across prompt versions, structured bug reports and QA automation. Through Codengage, I also created and trained AI models and built data products.

  2. Sep 2023 — Mar 2024

    Network Technician · Proveminas

    Cabling, hardware and internet connections. The hands-on infrastructure base behind everything after it.

  3. Feb 2026 — Jun 2030 (expected)

    B.A. International Relations · UNINTER

    Markets, globalization and regulation, the context software products operate in.

  4. Mar 2023 — Jul 2027 (expected)

    B.Sc. Software Engineering · UTFPR

    President of the CAES Alan Turing student council in 2025, secretary in the 2026 board. Research on LLMs for test-smell detection.

06 / Achievements & impact

Evidence, not adjectives.

  • 10+

    vendor prompt systems

    Authored or audited for the SubBase extraction pipeline (Wesco, Airgas, White Cap and others).

  • 500+

    automated tests shipped

    In subbase-qa-tests, tracked in MQ-Test, run on CI.

  • 98.85%

    extraction accuracy gain

    Measured across vendor prompt revisions in SubBase’s extraction pipeline.

  • Speaker, DevParaná na Estrada 2026 & Anhembi Morumbi (São Paulo)

    Talks on “From student to international developer” and “Paths and technology”: how early-career engineers reach global teams.

  • Mentor, IAthon (Sudotec / SENAI)

    Guided teams building AI prototypes under hackathon constraints.

  • Led the Software Engineering student council (CAES Alan Turing)

    President in 2025, secretary in 2026; organized the Semana Acadêmica and its sponsorships.

  • Research: LLMs for test-smell detection

    Comparative study of language models identifying test smells, written in SBC format at UTFPR.

07 / Contact

Let’s build software that works everywhere.

Organizing a talk, looking for a mentor or building a tech community? My inbox is open — English or Portuguese both work.

Photo of Herick Gonzaga

Herick Gonzaga

Beyond the code

Outside of work, I’m a big sports fan, a lover of Brazilian music in all its forms, and every now and then, someone who turns thoughts about life into words. I enjoy taking the stage to share ideas about technology and careers, helping people who are just starting their journey, and getting involved in volunteer projects with young people in my town. After all, I believe knowledge, experiences, and good stories are even better when shared. If you want to get to know me beyond GitHub, Instagram is the place.

@herick.andrey