YMS Agents and an early innovation award

In July 2025, I had the chance to present an initial prototype of agent workflows being explored in Micron for yield-related analysis. The work won the Innovation Award at the Data Science Technical Seminar for YMS Agents, which was encouraging because the project was still at a very early exploration stage.

I am grateful to be in a team that has the space to try this. Getting to explore early agents inside a real manufacturing environment is not something I take for granted. The tooling is still changing quickly, and there are not many established examples to copy from.

The models were not performing greatly yet. That meant the prototype had to stay practical and modest. We tried basic features first: retrieving relevant context, suggesting analysis paths, helping organize repeated investigation steps, and showing whether the agent could fit into the way engineers already think about yield issues.

• Keep the workflow understandable to engineers.
• Avoid building clever demos that could not survive real questions.
• Use the agent to reduce repeated exploration, not to replace engineering judgment.

The award helped give the idea internal momentum. More importantly, it taught me that AI work in manufacturing has to be explained through operational pressure, not model novelty. People want to know whether the tool helps them get to the next useful hypothesis faster.

Footnote: Ported over from my personal blog. Initially posted on 18 July 2025.