The business problem is to improve the accuracy and efficiency of Non-Destructive Testing (NDT) processes, specifically for flaw detection, segmentation, and classification in 3D volumetric CT data and 2D X-ray images, by leveraging AI and machine learning models integrated into real-time production line inspection systems.
Requirements
- Experiences classical image processing techniques for NDT (e.g., filtering, edge detection, thresholding)used in conjunction with ML
- Experience with point cloud processing and 3D geometry analysis.
- AI Model Development: Design, train, and optimize deep learning models (e.g., 3D CNNs, Vision Transformers, U-Nets) for flaw detection, segmentation, and classification in 3D volumetric CT data and 2D X-ray images.
- Algorithm Validation: Develop rigorous validation frameworks to measure model performance against expert NDT technician analysis.
- Data Pipeline Architecture: Build robust data pipelines for ingesting, preprocessing, and augmenting large-scale volumetric CT data.
- System Integration & Deployment: Work closely with automation and robotics engineers to integrate models into real-time production line inspection systems.
- Optimize models for high-speed inference and edge deployment.
Responsibilities
- Design, train, and optimize deep learning models (e.g., 3D CNNs, Vision Transformers, U-Nets) for flaw detection, segmentation, and classification in 3D volumetric CT data and 2D X-ray images.
- Develop rigorous validation frameworks to measure model performance against expert NDT technician analysis.
- Build robust data pipelines for ingesting, preprocessing, and augmenting large-scale volumetric CT data.
- Handle challenges like noise reduction, normalization, and material attenuation artifacts.
- Work closely with automation and robotics engineers to integrate models into real-time production line inspection systems.
- Optimize models for high-speed inference and edge deployment.
- Monitor model performance in production and implement active learning and data drift detection strategies to continuously improve system accuracy.
Other
- Understanding of industry standards and regulations (e.g., NADCAP, AS9100, ISO 9001) related to quality control and NDT processes.
- Familiarity with robotics, PLCs, and industrial automation systems (e.g., Siemens, Fanuc).
- Cross-Functional Collaboration: Partner with NDT Level II/III technicians to understand defectology and define accurate labeling protocols.
- Work with mechanical and materials engineers to incorporate domain knowledge (e.g., material properties, expected stress points) into model design.
- MASTER OF COMPUTER SCIENCE
