Selected work

Scale AI · Project manager, AI safety

Scaling contributor operations

Our contributor team went from fewer than 20 to more than 200 as incoming project volume tripled. I helped build the qualification and progression framework and made QA reporting consistent for project leads.

When is someone ready for the next level?

With more projects arriving, leads needed a consistent way to decide who could take on more complex work. A high score from a handful of tasks gave us limited evidence. We looked at the quality score alongside a minimum volume of reviewed work. The framework included a way for reviewers to revisit disputed scores and agree on how to apply the criteria.

What happens when someone needs more practice?

The progression plan gave contributors a route back through targeted retraining. I documented how leads could group people by the errors showing up in their work, such as interpreting guidelines, and use task buddies, office hours and individual feedback to address those gaps.

For contributors moving into the advanced stage, the plan called for a week doing the work and a second week learning to review it. Leads could use the two weeks of work to decide whether the contributor would stay in that role.

What project leads needed to see

I standardized QA reporting around audit findings, throughput, bottlenecks, and escalations. Leads could see which issues called for calibration, more training, or attention to delivery. The operating framework also called for a shared reference for guidelines, project scope, and deadlines across the distributed team.