← Back

OptiViz: AI-Powered Steel Surface Quality Control System

5/10/2025

Led the development of a computer vision system that automates steel surface defect detection across manufacturing lines. Combined deep learning, microservices, and real-time visualization to transform quality control.

Technical highlights

  • Custom 3-layer CNN (DefectNet) in PyTorch identifying six critical defects: crazing, inclusion, patches, pitted surface, rolled-in scale, scratches
  • Scalable microservices with Docker; FastAPI for high-performance inference endpoints
  • Apache Airflow for automated data pipelines and retraining orchestration
  • React dashboard with D3.js visualizations for real-time defect tracking and confidence scores
  • Continuous data ingestion loop for ongoing model improvement

Impact

  • Streamlined quality control via automated detection
  • Reduced manual inspection time with live visualization
  • Improved detection accuracy through continuous retraining
  • Boosted efficiency with automated QA processes

Tech stack Python, PyTorch, FastAPI, React, D3.js, Docker, Apache Airflow, Computer Vision, Deep Learning, Microservices Architecture

Computer VisionPyTorchFastAPIReactD3.jsDockerAirflowMicroservices