Apollo Dash

In June 2022, 9 months before the NovaSeq X product launch, end-to-end yield for Flow Cell Manufacturing was sitting at 94.26% - but 97.5% was needed to reach per-unit cost targets. That gap looks small - but closing it meant preventing more than 55% of all scrap on a line with hundreds of distinct failure modes.

I was recruited internally to design and implement a data strategy for the new line. The scope was open-ended - previous tools on legacy lines had failed to gain traction, and my starting point was to find out why: what was and wasn't working, and for whom.

I was the sole PM on this project, without direct authority over any of the engineering resources. I had to advocate for and coordinate contributions from data engineering, data science, and product engineering, while working closely with quality, validation, and process engineering throughout.

Yield rose from 94.26% to 97.98%. $14M in annual cost savings at early production volumes - a figure expected to grow as NovaSeq X volume scales - and 100% tool adoption across the team. Perhaps more durably, the team shifted from reactive firefighting to proactive troubleshooting - a change in how engineers relate to data that no metric fully captures.

The impact of any user-facing data tool depends on the action a user takes based on the information they receive. The foundation for user action is trust.

I defined the scope, aligned the team, stakeholders, and users around it, and drove execution.