Problem / System
Advanced semiconductor manufacturing requires identifying extremely small process and alignment errors, often through indirect measurements collected under severe throughput and instrumentation constraints.
My Work
Developed statistical, physical, and machine learning approaches for metrology and manufacturing problems involving imaging, process variation, overlay, edge placement, and measurement system behavior.
Technical Challenge
The work involved inferring manufacturing behavior from imperfect, indirect, and differently sampled signals while keeping model outputs useful within real production workflows.
Public Artifacts
A patent on machine learning for synergistic metrology is the main public technical artifact.
While proprietary details are not reproduced here, ASML’s Investor Day materials show the Applications roadmap that encompassed my work. The business line was projected to grow by ~20% annually, with the Virtual Computing Platform playing a central role. This is company-level reporting, but it gives a sense of the broader opportunity described below. I played a critical role in establishing an Intel foothold for that opportunity.
Impact
I turned a metrology product that had failed to gain customer traction into a viable Intel offering by finding an adoption path that fit their existing process control workflow, building a first-principles prototype that exposed flaws in the algorithm and motivated a key design pivot, and proving its value in real dispositioning decisions. This created an initial $30M licensing opportunity and, by introducing new compute requirements, opened a broader platform opportunity exceeding $300M that I continued to shape through technical recommendations and customer-facing product work.
In separate work, I developed a patented ML-based method for inferring higher-resolution measurements of nanoscale chip features from lower-resolution optical signals.