Master in Electrical Engineering
During my internship as a machine learning engineer in semiconductor manufacturing, I worked on detecting harmful defects from images. The challenge was harder than expected due to subtle defects, imbalanced data, and labeling issues. I started with a pretrained EfficientNet-B0 model to set a baseline, then experimented with different architectures like convolutional networks and vision transformers. I also used adaptive gradual layer unfreezing and data augmentation to improve performance. Handling 40,000 unlabeled images, I tried self-supervised learning and clustering to find patterns and reduce manual labeling. Beyond technical work, I learned to communicate and collaborate across hardware, software, and QA teams. My mentor encouraged independent thinking and thorough documentation, helping me build more robust solutions. Working in a multicultural environment improved my collaboration skills. This internship taught me to balance accuracy with practical needs, adapt quickly, and communicate clearly, these lessons that will guide me in future machine learning projects.
