Systems Analyst at WEG Energia · Control & Automation Engineer
I work on the engineering systems behind large electric machines. At WEG Energia, the WEG division that builds engineered-to-order motors and generators, I develop and maintain the SAP-based system that generates engineering bills of materials from configuration rules.
My path to that role was not a straight line: enterprise software since 2012, a control and automation engineering degree from UFSC, and hands-on industrial work — PLC communication, machine retrofits, machine learning applied to real equipment.
WEG is the world's largest manufacturer of low-voltage electric motors — it overtook ABB in 2024 — with plants in 17 countries. WEG Energia (WEN) is the unit that builds engineered machines: large motors and generators designed one order at a time, each with its own bill of materials running to thousands of items.
I'm part of the Technology Development & Innovation department inside WEG Energia's engineering directorate. My work is the system that turns engineering rules into manufacturing data: a rule-based BOM generator built inside SAP. Engineers describe a machine's configuration; the rules engine resolves which materials go into which BOM positions, with validity rules tied to machine IDs, material IDs, rule groups and item-level conditions.
Day to day that means SAP (ECC, S/4HANA, NetWeaver Java, Solution Manager), databases from Oracle and SAP HANA to SQL Server, PostgreSQL and MongoDB, data modeling in service of engineering, and reading SolidWorks drawings of machines and motor parts to understand what the data has to represent. It sits in the same problem space as SAP PLM and variant configuration for engineer-to-order manufacturing.
Development, maintenance and engineering support of the SAP-based BOM automation system for engineered-to-order electric machines. Data relations, configuration rules, technical lists and specifications.
Retrofit of legacy industrial machines. Built a web HMI on a Raspberry Pi talking Modbus RTU over RS485 to the machine's PLC — still running in production.
Predictive maintenance research for industrial machines. Edge AI with Google Coral TPU, Node-RED and MQTT sensor pipelines, MLOps deployment with GitHub Actions, Terraform and AWS.
Corporate travel-expense SaaS: Python/Django services, REST APIs, PostgreSQL, ETL pipelines and AWS infrastructure.
ERP modules for textile manufacturing (Delphi XE, Oracle PL/SQL), used daily on the factory floor.
Thermal cameras plus convolutional networks to catch three-phase motor failures before they stop a line. MobileNetV2 with transfer learning reached 0.912 ROC AUC on a custom thermography dataset; Grad-CAM heatmaps show maintenance teams where the model is looking. Deployed on an NVIDIA Jetson Xavier with OPC UA and MQTT for SCADA integration.
UFSC repository TensorFlow · Grad-CAM · Jetson · OPC UA
The original operator panel of a Neocoude HD-15 died; a replacement cost more than the machine was worth. I put a Raspberry Pi next to the PLC, spoke Modbus RTU at 57,600 baud over RS485, and served the controls as a web page — any tablet on the shop network became the HMI. Python asyncio backend, WebSocket updates, a 400 PPR encoder for position tracking and a 7-step automatic calibration routine.
GitHub Modbus RTU · RS485 · asyncio · in production
Object detection at 30 FPS entirely on-premises: TensorFlow models compiled for the Google Coral Edge TPU, Frigate NVR, MQTT and ESP32 sensors, all containerized. No cloud, no footage leaving the building.
GitHub TFLite · Coral TPU · Docker · MQTT
A commit-to-production path for ML services: pytest, Docker build, and Terraform-provisioned AWS deployment on every push. Written to prove the boring parts — the ones that keep a model running at 3 a.m. — not the model itself.
GitHub Terraform · GitHub Actions · AWS
Outside my role at WEG I keep a small consulting practice through my own companies, Blumenau TI and Blumenau Automação (CNPJ 53.700.754/0001-10). I take on selected B2B projects — machine retrofits, hardware-to-software integration, and automation tooling — and can invoice international clients directly (W-8BEN-E available).
If you have an industrial machine that needs to talk to modern software, or engineering data that needs untangling, write me. I answer everything sent to the address below.