Alexandre Augusto Arduini Tavares

Design hardware and software architectures for industrial computer vision using open-source tools. Lead data acquisition, labeling, model training and deployment across concurrent projects.

• Design hardware and software architectures for industrial vision systems
• Lead acquisition, labeling, model training and deployment
• Manage schedules and cross-functional collaboration across concurrent projects

Developed open-source defect-detection and quality-assessment algorithms. Built a Python algorithm using OpenCV and PIL to identify millimetric surface defects in high-resolution coil imagery, rolled out across multiple plants.

• Built a Python, OpenCV and PIL algorithm for millimetric surface-defect detection in high-resolution coil imagery
• Rolled out the surface-defect inspection solution across multiple plants, improving inspection reliability
• Mentored junior engineers and identified computer vision opportunities for the innovation team

Initiated the company's first predictive-maintenance data-science project. Installed and monitored more than 1,000 sensors on more than 100 assets, and helped establish a central monitoring room.

• Initiated a predictive-maintenance data-science project
• Installed and monitored more than 1,000 sensors across more than 100 assets
• Applied data-driven prioritization under tight budget constraints

Deployed a trunked radio system serving more than 1,000 employees. Installed and maintained more than 500 security cameras, and maintained routers and servers.

• Deployed trunked radio serving more than 1,000 employees
• Installed and maintained more than 500 security cameras
• Modernized radio communications within budget through cross-team collaboration

Developer at RCS Informática from approximately December 2013 through May 2014.

Engineering Intern at Aperam South America from September 2013 through November 2013.

Numerical Calculation Academic Monitor at Universidade Federal de São João del-Rei (UFSJ) from February 2013 through August 2013.

Improved an existing Python vision inspection system through processing corrections, profile-specific calibration and web inspection views. Evaluated changes using archived images; production validation remains pending.

Developed a Python, OpenCV and PIL algorithm for millimetric surface-defect detection in high-resolution coil imagery. Rolled out across multiple plants, improving inspection reliability.

Rebuilding an AI-assisted workflow to reduce manual tag-verification effort in complex brownfield P&IDs. Own architecture, cloud infrastructure, backend, frontend and model integration, connecting extraction, engineer correction and reviewed inventory exports. Time savings and error reduction remain to be validated.

Built an independent Python CLI and local dashboard for planning AI agent work, with approval checkpoints, rule-based task routing and records linking outputs to their inputs.