Skip to content
View mdoin's full-sized avatar

Block or report mdoin

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
mdoin/README.md

Hi, I'm Maitê

Software engineer in Londrina, Brazil. I build production systems in Python, Node.js and PostgreSQL, with AI coding agents in the loop every day.

Self-taught. I started by automating my own work at an insurance billing desk and kept going until it became the job.

What I work on

  • Computer vision on outdoor media: real-time pipelines over camera streams that count people and vehicles and measure attention on LED panels, without storing personal data.
  • Multi-tenant SaaS: a CRM pulled out of a monolith and turned into a product, with per-tenant schemas, a home-grown automation engine and live updates.
  • AI tooling: custom MCP servers that expose internal systems (PostgreSQL, ERP APIs, an LED media panel) as agent tools, with scoped permissions and audit logs. Local models via Ollama when data can't leave the network.
  • Internal tools: 10+ in the last six months, mostly replacing spreadsheet-and-copy-paste workflows for operations teams.

Selected projects

ExpoTimeOut: live inference on outdoor LED media
Counts people and vehicles, estimates age range and attention time, and runs an activation where visitors generate AI artwork that reaches the panel via QR code in under a minute. The hardest part was infrastructure, not models: cameras with no reachable RTSP endpoint, solved with RTMP push ingest.
Python YOLO ByteTrack InsightFace FastAPI PostgreSQL

LED panel at an expo booth showing a live count of people in front of it, average age, and vehicle impressions
Live audience count on the panel at an expo booth.

Multi-tenant CRM
Schema-per-tenant isolation in PostgreSQL, an automation engine with an expression parser written from scratch (no eval), full-text search, and real-time updates over SSE with LISTEN/NOTIFY.
Node.js Express PostgreSQL React Docker Caddy

AI financial copilot
A chat agent that answers financial questions from real ERP data and acts through function calling: reconciling invoices, drafting contracts. Nothing financial runs without human approval plus 2FA, and every action lands in an append-only audit log.
Fastify React RAG OAuth2 ETL

Stack

Python, TypeScript, Node.js, FastAPI, PostgreSQL, React, Docker, Linux

Plus YOLO, InsightFace, Claude Code, MCP and Ollama on the vision and AI side.

Now

  • Software engineer at Intermidia
  • Technologist degree in Systems Analysis & Development at Estácio
  • Running Doin Labs: landing pages, automations and applied AI for small businesses

Contact

LinkedIn · maitedoin@gmail.com

Pinned Loading

  1. contadorooh contadorooh Public

    Real-time computer vision that measures how many vehicles pass a street LED panel, what type they are, how long they are exposed to it, and how many people that represents.

  2. copilotofinanceiro copilotofinanceiro Public

    Internal finance automation (receivables, payables, bank reconciliation, collections and natural-language queries) that treats AI as a supervised assistant, never as a financial authority.

  3. crmsaas crmsaas Public

    Multi-tenant web app combining sales CRM, WhatsApp support and bulk email, with per-tenant isolation, per-account branding and an internal operations console.

  4. interativooh interativooh Public

    Out-of-home media system (LED panels) that generates a custom ad from a mobile form, moderates it automatically and puts it on air in minutes, with proof of play sent back.