01
Optical tomography
Use imaging captured during laser powder bed fusion to study process behavior and indications of defects.
PhD Research / University of Waterloo / 2020–2024
Industrial Computer Vision
Finding defects in metal additive manufacturing through optical tomography and machine learning.
My doctoral research investigated how imaging collected during laser powder bed fusion can support anomaly detection and quality assessment. I developed ML pipelines and segmentation approaches to turn complex monitoring data into useful defect information.
Optical TomographyU-NetSelf-Organizing MapsFuzzy LogicDefect Detection
Read the publications The problem
Metal parts can develop internal defects during printing. Post-production inspection helps assess quality, but monitoring the build itself offers an opportunity to identify indications of defects earlier.
Optical tomography captures light emissions from the process. Interpreting those images is challenging: process variation and imaging disturbances can obscure the patterns associated with porosity.
My contribution
From monitoring images to defect information
My work spanned imaging-data pipelines, model development, and evaluation. Across the publications, I investigated complementary ways to learn from optical tomography data and incorporate manufacturing knowledge.
02
Self-organizing maps
Explore patterns in monitoring data to support the identification of process conditions associated with defects.
03
U-Net segmentation
Develop image segmentation models to localize defect-related regions in optical tomography images.
04
Fuzzy logic
Incorporate domain knowledge through adjustable rules, connecting learned patterns with interpretable quality decisions.
Evaluation & outcomes
Connecting predictions to physical defects
The 2024 study evaluated predictions of lack-of-fusion and keyhole defects against subsequent CT scans. It also investigated configurable fuzzy rules and probability thresholds, allowing quality requirements to influence the detection decision.
This work contributed to two journal publications. The linked papers document the experimental conditions, evaluation measures, and results. Performance depends on the process conditions and evaluation setup.
What I bring to AI engineering
Experience working with complex imaging data, developing segmentation models, and evaluating predictions against physical measurements. This is the foundation I bring to building and assessing practical ML systems.
Research outputs
Selected publications
Additive Manufacturing / 2023
Optical tomography and machine learning for in-situ defects detection in laser powder bed fusion: A self-organizing map and U-Net based approach
International Journal of Extreme Manufacturing / 2024