Work Activities
As electronic chips continue to shrink while becoming increasingly powerful, their fabrication depends on achieving sub-nanometer precision at high speed and efficiency. At the heart of this technological challenge lies the need for accurate and reliable metrology guided by cutting-edge computational methods.
Recent breakthroughs in Machine Learning enable powerful physics-informed models capable of solving the inverse problem of reconstructing metrology parameters directly from low-resolution microscopy images. By embedding this problem in a Bayesian framework, these models not only estimate the desired physical quantities but also provide meaningful uncertainty quantification. Such uncertainty estimates are crucial for both process control in semiconductor manufacturing and broader scientific applications.
In this project, you will develop a novel Machine Learning approach that integrates physical simulations of the measurement process with its inve...