Holland Smith

I work on applied AI and optimization for semiconductor manufacturing. I'm currently Director of Data Science at INFICON, where I build LLM-based and multi-agent systems that help engineers in fabs around the world diagnose production issues and make better decisions.

Background

I started out in physics and materials science, doing my Ph.D. at UC Berkeley and Lawrence Berkeley National Lab on semiconductor materials for radiation detection. From there I moved into industry as a process engineer at Intel, then spent several years deploying scheduling, capacity modeling, and analytics software in factories worldwide, first at Final Phase Systems and then at INFICON.

Along the way my work shifted from the physics of the devices to the systems that run the factories that make them, and more recently to the AI tools that help people understand and operate those systems. A lot of my time goes into working directly with the teams who use this software, from engineers on the floor to leadership, figuring out what problem is actually worth solving.

Interests

• Agentic AI, retrieval-augmented systems, and how to evaluate them honestly
• Scheduling and combinatorial optimization, especially flexible job shop problems
• Materials science, semiconductor physics, and solid state physics
• Semiconductor manufacturing and the data it generates
• Explaining complex technology clearly, to whoever needs to hear it
• Languages and literature (I also have a B.A. in Slavic Languages from Stanford)

Talks & writing

I've spoken on agentic AI and manufacturing at SEMICON West, SEMICON Europa, ASMC, APCm Europe, and AEC/APC Japan, and have published papers in applied physics, industrial engineering, applied machine learning, and semiconductor materials.

Contact: holland@hollandsmith.com · LinkedIn