Graph-based detection infrastructure and reviewer tooling that protected the ad exchange ecosystem at global scale.
Vikram built detection infrastructure that empowered 200+ fraud analysts to identify millions of abusive accounts — converting individual review into coordinated, ring-level intelligence.
The ad exchange ecosystem faced coordinated abuse at massive scale — fraud networks operating across millions of accounts with behavioral patterns that individual account review could never catch. Manual review processes were reactive and slow, unable to surface the ring-and-chain relationships that distinguished coordinated fraud from isolated violations.
The detection challenge was compounded by the scale of data involved: processing massive volumes of account behavioral data in near-real time while maintaining accuracy high enough to avoid penalizing legitimate advertisers. Continuously evolving fraud patterns required the system to be architecturally adaptable, not just accurate at point-in-time.
"The goal wasn't to catch more fraud — it was to give 200 human reviewers the intelligence to catch fraud that no single person could ever see alone."
The platform identified millions of abusive accounts, strengthening the integrity of the ad exchange advertising ecosystem at global scale. By empowering 200+ reviewers with intelligent tooling — graph visualization, similarity scoring, and bulk actions — the system multiplied human reviewer capacity without adding headcount. The project earned a VP Award, reflecting its significant financial and operational impact on one of the company's most revenue-critical platforms.