Ad Operations Engine
Automating item categorization and validation with ML
Identified operational bottlenecks and led the development of an ML-powered tool that automated item categorization and validation, reducing manual effort and helping customers get their items live faster.
Problem / Context
All item tagging, validation, and approval workflows were handled manually, creating severe operational bottlenecks, delaying item go-live speeds, and frustrating advertisers waiting for campaigns to launch.
Goal & Success Metrics
Go-Live Speed
5× faster item time-to-live
Operational Overhead
Drastic reduction in manual review bottlenecks
Options, Decisions & Trade-offs
First built a lightweight version without integrating with the entire system. Worked alongside the operations team for a few weeks to validate that the model performed reliably on edge cases, and only after confirming high accuracy did we integrate it into the core pipeline through iterative releases.
Learnings & Retrospective
What worked well
Influencing stakeholders without formal authority, building trust with the operations team to adopt ML automation.
What I would do differently
Bring operational stakeholders into problem discovery and solution design much earlier in the process.