I help companies integrate legacy industrial hardware with modern software, computer vision, and AI —
delivering complete solutions from factory-floor PLCs to cloud-deployed ML models.
Available for remote B2B contracts (US / Global).
End-to-end automation and AI engineering for companies that need results — not excuses
I'm an Automation & Machine Learning Engineer with a B.Eng. in Control & Automation from UFSC (Brazil) and a background that spans industrial hardware, embedded systems, full-stack software, and production AI deployment. I manage projects end-to-end and lead a small engineering team — from initial architecture to final handoff.
My niche is the gap most companies struggle to fill: connecting legacy industrial hardware with modern software and AI. I've built real-time HMIs talking to PLCs via Modbus RTU, deployed computer vision models on edge devices, and wired MLOps pipelines from scratch. Whether you need a quick proof-of-concept or a production-grade system, I deliver.
I operate through my incorporated Brazilian IT company (CNPJ 53.700.754/0001-10), making international B2B engagements simple and compliance-friendly. My Italian passport means I can also work on-site anywhere in the EU without restrictions, or travel to US clients for critical project phases under the Visa Waiver Program.
I handle the entire stack — PLC to cloud, hardware to API. No need to coordinate 5 vendors.
I build systems that run in factories, not just Jupyter notebooks. Real machines, real ROI.
B2B contract via incorporated company — no W-2, no payroll tax, no benefits admin. Clean invoicing.
I can work solo or bring my engineering team — from a one-week consulting sprint to a multi-month project.
Full-stack engineering from factory floor to cloud — I cover the stack most teams leave as gaps
Production systems delivering real business impact — hardware to cloud
Industrial AI system that detects failures in 3-phase induction motors using thermal cameras and deep learning — before failures cause costly unplanned downtime. Grad-CAM heatmaps make AI decisions visually interpretable for maintenance teams, enabling trust and fast corrective action on the factory floor. Deployed on edge hardware for real-time inference and integrated with OPC UA/MQTT for SCADA connectivity.
Predictive maintenance replaces reactive repair cycles. By detecting thermal anomalies hours or days before motor failure, this system enables maintenance scheduling during planned stops — eliminating unplanned production halts, cutting spare parts costs, and extending equipment lifespan. Grad-CAM visual feedback accelerates maintenance team decision-making and builds operator trust in AI systems.
Full modernization of a 15 HP industrial hydraulic rebar bender (Neocoude HD-15, capacity up to ⌀50 mm rebar). The original HMI panel was damaged — hardware replacement would have cost €2,000+. Instead, I bridged the gap between raw hardware and user-friendly software by building a modern Web HMI running on a Raspberry Pi 3B+, communicating directly with the machine's PLC via Modbus RTU over RS485. Any tablet or smartphone on the local network becomes the operator interface.
Eliminated a €2,000+ hardware replacement cost while adding capabilities the original panel never had: remote monitoring, automatic 7-step calibration, production data logging, and Tailscale VPN for off-site access. This is the ROI that makes retrofit engagements compelling: spend less, get significantly more.
Custom Keyboard, Video, and Mouse (KVM) emulation architecture for remotely controlling multiple physical machines over the network. Developed a scalable distributed system that routes keyboard and mouse inputs from a central Raspberry Pi server to multiple target computers via Raspberry Pi Pico W2 client nodes — each appearing as a native USB HID device, invisible to the host OS, requiring zero drivers.
Replaces commercial KVM switches ($200–$1,000+) with a fully custom, extensible solution. Ideal for industrial labs and manufacturing environments where managing multiple embedded computers is routine and physical access is impractical.
Production-grade anomaly detection service that proves I don't just train models — I deploy them and keep them running. A FastAPI service serving ML inference with Pydantic validation, automated pytest coverage, Docker containerization, and a full CI/CD pipeline. Designed for industrial sensor data streams where reliable, low-latency detection is non-negotiable.
Most ML engineers hand off a model file. This is a deployable, maintainable, testable service — the kind that stays running at 3 AM without paging your on-call team.
Production-ready AI-powered surveillance system with Google Coral Edge TPU hardware acceleration. Real-time object detection at 30 FPS with <100ms latency — fully local processing, zero cloud dependency, GDPR-compliant by design. Complete ML lifecycle from data collection to containerized deployment.
Production-grade MLOps pipeline with Infrastructure as Code. Every commit triggers automated tests, Docker build, and Terraform-provisioned AWS deployment. Demonstrates the DevOps discipline required to run ML in production responsibly — not just once, but repeatably.
Flexible engagement models designed for international B2B clients. From a quick consulting sprint to a full project.
From factory-floor PLCs and microcontrollers to cloud-deployed Machine Learning models — I cover the full engineering stack. No need to coordinate multiple vendors or leave the gap between your hardware team and your data science team unfilled. I am that gap.
I operate as an independent contractor through my incorporated IT company in Brazil (CNPJ 53.700.754/0001-10). Simple, compliance-friendly B2B agreements — W-8BEN-E ready. No payroll tax, no benefits administration, no HR overhead for your US entity.
Experienced integrating with global engineering teams across time zones. I deliver well-documented code, clear async communication, and reliable sprint commitments. Italian passport provides the flexibility to work on-site anywhere in the EU or visit US clients for critical project phases — no visa friction.
Connect legacy PLCs, sensors, and industrial machines to modern software. Modbus RTU/TCP, OPC UA, RS485, custom protocols — hardware that talks to everything.
Deploy thermal imaging, vibration analysis, or sensor fusion ML models to predict equipment failures before they cause downtime. End-to-end: data collection → model → edge deployment.
Defect detection, object recognition, thermal analysis — from model training to edge deployment on Jetson, Coral, or Raspberry Pi. Fast, accurate, runs where you need it.
Modern browser-based operator interfaces replacing aging HMI panels. Real-time WebSocket communication, tablet/mobile-friendly, remote VPN access. No proprietary HMI software required.
FastAPI services, Docker containers, CI/CD pipelines, automated testing — production-grade ML infrastructure built to stay running. Not a research prototype, a product.
Architecture review, technology selection, project scoping, and feasibility analysis for automation and AI initiatives. Hourly, retainer, or one-off engagements — no minimum commitment.
Federal University of Santa Catarina (UFSC) — Blumenau Campus
Blumenau, Santa Catarina, Brazil
Thesis: Predictive Failure Detection in Three-Phase Induction Motors Using Thermography and CNNs with Grad-CAM Interpretability
Key Coursework: Control Systems, Industrial Automation, Embedded Systems, Signal Processing, Instrumentation, Machine Learning
Complete Technical Formation
Focus Areas: Programming, Databases, Networking, Systems Administration, Software Development
Available for remote B2B contracts · US, EU & global clients welcome · Response within 24 hours
Incorporated Brazilian IT Company
CNPJ 53.700.754/0001-10
W-8BEN-E available on request
Blumenau, Brazil
Remote-first · EU on-site available (Italian passport)