Arthur Emmanuel.Consulting & Engineering
PT/ENLet’s talk
Arthur Emmanuel, software and AI engineer in Lagoa Santa, Brazil
Arthur Emmanuel · Lagoa Santa, Minas Gerais, Brazil
About · Arthur Emmanuel

Software built with AI agents. And with method.

Software engineer focused on agentic AI. I combine specifications, AI agents and process analysis to deliver systems that work day to day — faster, without giving up quality.

GuideAI assistant · replies now

So I can show you what matters to you:

What brought you here?

Private conversation. You decide what reaches Arthur.

Talk to ArthurSee how I build
  • Agentic AI
  • Spec-driven development
  • Process analysis
Scroll to meet him
02The origin

Before writing code, I saw where work gets stuck.

I worked inside an administrative operation before building systems. It became clear that the problem is rarely a missing tool: it’s information that never arrives, tasks that depend on someone remembering, decisions made without numbers.

That’s why I don’t start with a technology. I start with a question — what needs to work better? Engineering comes next. And today it comes with artificial intelligence.

  1. 01Understand before buildingThe process diagnosis comes before the first line of code.
  2. 02Written before doneEvery rule and acceptance criterion is written down, for people and agents.
  3. 03AI where it saves timeAnd only there. Technology is a means, not a showcase.
03How I build

Seven disciplines. One way of working.

Tap each point of the constellation. They all connect to the one in the centre: agentic AI.

Agentic AI

What it is
AI agents that plan, use tools and carry out tasks in steps — not just answer questions.
How I apply it
Agent orchestration with tool calls, closed actions and safety limits. In development, coding agents work from specs and versioned instructions, with human review.
What you gain
Work that used to need someone watching over it now happens on its own — with oversight where it matters.
04Toolbox

Tools change. The method stays.

What I use today to build with AI agents — and what each group is for.

01

Agentic development

To build faster without losing control of what gets delivered.

  • Claude Code
  • Google Antigravity
  • AGENTS.md and versioned skills
  • Specs as the source of truth
  • Human review at every stage
02

AI models and APIs

To put intelligence inside the process, with verifiable answers.

  • Gemini API
  • Gemini Live (real-time voice)
  • Function calling
  • Structured outputs
  • Multimodal models
03

Software engineering

For systems that are fast, secure and grow with the business.

  • Next.js
  • React
  • TypeScript
  • Node.js
  • Firebase
  • Git and GitHub
04

Quality, data and integrations

To prove it works — and connect what already exists.

  • Playwright
  • Zod
  • Redis
  • Meta Graph API
  • APIs and webhooks
05Trajectory

Every stage added a layer to what I deliver.

  1. Computer Engineering

    UFOP · previous studies

    What it gave me: The foundation: algorithms, systems and the habit of understanding how things work inside.

  2. Administrative assistant

    Unimed João Monlevade

    What it gave me: Operations seen from the inside: processes, queues and rework — the problem before the software.

  3. Tech Lead · Architecture · Full-stack · UI/UX

    Digital ecosystem

    What it gave me: An entire ecosystem: from the public website to the internal platform, with AI in production.

    See the KROFH case →

  4. Information Systems

    PUC Minas · in progress

    What it gave me: Management, data and information systems applied to business.

  5. Project management & AI research

    WebTech Network · VUEI / PUC Tec

    What it gave me: Autonomous agents and machine learning in real research projects.

Certifications

  • JavaScript: objects
  • Foundations of Cybersecurity
  • HTTP: understanding the web under the hood
  • HTML5 and CSS3: positioning, lists and navigation
  • From scratch: programming for beginners
06Why me

You talk to the person who builds. From diagnostic to delivery.

  1. 01One person, start to finish

    Whoever runs the diagnostic is who specifies and builds. No middlemen, no broken telephone.

  2. 02AI speed, engineering rigour

    Agents speed up construction; specification and tests guarantee the result.

  3. 03Operations understood from inside

    Experience in an administrative and operational environment: I know how a real routine works.

Book a diagnostic
07The decision

If it made sense so far, the next step is a conversation.

I don’t know yet what brought you here — that’s fine. The conversation starts wherever you like.

You are
Choose
What interested you most
Choose
Next step
Choose

Shall I take this summary to Arthur?

Continue on WhatsAppBook a diagnostic
LinkedIn GitHub

A direct conversation with Arthur. You review the message before sending.

Arthur Emmanuel Oliveira Marinho

Software engineer · Agentic AI, automation and systems

arthuremmanuel.dev@gmail.com · +55 31 99068-1099 · Lagoa Santa, Minas Gerais, Brazil
https://www.linkedin.com/in/arthur-emmanuel-oliveira-marinho/
https://github.com/arthuremmanuel01

I build systems, automations and AI agents with specifications, tests and architecture. Experience with processes and project management, currently doing research with autonomous agents and machine learning.

Competencies

  • Spec-Driven Development. The spec is the source of truth for people and for AI agents. Every delivery is checked against it, with automated tests.
  • Prompt and context engineering. Curated knowledge bases, versioned system instructions and structured outputs validated by schema.
  • AI modelling and evaluation. Classification, extraction and summarisation with LLMs, retrieval (RAG), multimodal models and machine-learning research.
  • Process analysis. Current and target flows, bottlenecks, owners and indicators — before choosing any technology.
  • Process automation. APIs, webhooks and integrations with WhatsApp, CRM, ERP and spreadsheets, with scheduled functions and alerts.
  • Software architecture and full-stack. Next.js, React, TypeScript, Node.js and Firebase; multi-tenant architecture, role-based access and end-to-end tests.
  • Agentic AI. Agent orchestration with tool calls, closed actions and safety limits. In development, coding agents work from specs and versioned instructions, with human review.

Experience

  • Project management & AI research — WebTech Network · VUEI / PUC Tec (2026 — present). Autonomous agents and machine learning in real research projects.
  • Tech Lead · Architecture · Full-stack · UI/UX — Digital ecosystem (KROFH). An entire ecosystem: from the public website to the internal platform, with AI in production.
  • Administrative assistant — Unimed João Monlevade (2024 — 2025). Operations seen from the inside: processes, queues and rework — the problem before the software.

Education

  • Information Systems — PUC Minas · in progress (2025 — 2029)
  • Computer Engineering — UFOP · previous studies (2022 — 2025)

Tools

Claude Code · Google Antigravity · AGENTS.md and versioned skills · Specs as the source of truth · Human review at every stage · Gemini API · Gemini Live (real-time voice) · Function calling · Structured outputs · Multimodal models · Next.js · React · TypeScript · Node.js · Firebase · Git and GitHub · Playwright · Zod · Redis · Meta Graph API · APIs and webhooks

Certifications

JavaScript: objects · Foundations of Cybersecurity · HTTP: understanding the web under the hood · HTML5 and CSS3: positioning, lists and navigation · From scratch: programming for beginners