Oct 2022 – Apr 2025 · 2 yrs 6 mos · On-site

• Used Gemini (Google's LLM) to boost upstream product defense by improving approval checks done during account sign-ups. Accounts with fake US addresses and/or drastic changes in content ("food recipes" to "education loans") were flagged for review.
• Streamlined multiple operational workflows (customer ticket routing form, user report form) to improve user experience and reduce the “time to completion” for various internal forms and processes.
• Discovered and mitigated novel abuse patterns involving false social media advertising and payment instrument abuse, worth ~$45M/year of revenue.
• Trained new team hires, completing their onboarding within 2 months of joining, including delegating training sessions, consistent weekly syncs and knowledge transfer sessions.

Jul 2017 – Oct 2022 · 5 yrs 3 mos · On-site

• Launched multiple classifiers - using SQL and Python - to launch data pipelines to flag and enforce websites for policy abuse.
• Managed Child Safety workflow for Google AdSense - re-trained vendors to reduce false positives, re-wrote enforcement guidelines & policy language to clarify gray areas for publishers.
• Improved performance of classifiers to flag policy abuse in non-english languages by leveraging vendor's knowledge of local languages, to re-train models with new keywords more relevant to the abusive content.
• Built dashboards to report enforcement reasons. Consumed by appeal reviewers, as well as Account Managers to help remediate their affected clients.