London, England, United Kingdom
As a Software Engineer at Meta, I contribute to the Ads Targeting & Audience Management team by building backend systems that integrate machine learning, distributed infrastructure, and advertiser growth across Facebook, Instagram, and Audience Network. My work includes developing an ML intent classifier using LLM embeddings to predict new-customer-acquisition campaigns and designing audience infrastructure for a messaging platform launch, enabling significant adoption and uplift in business uploads.
Ads Targeting & Audience Management — building backend systems at the intersection of ML, distributed infrastructure, and advertiser growth across Facebook, Instagram, and Audience Network (~1B+ API requests/week). Impact highlights: 🤖 Built an ML intent classifier using LLM embeddings (85% recall / 90% precision) to predict new-customer-acquisition campaigns — enabling automated audience expansion without advertiser manual setup 📈 Drove 1B+ unique subscribers and $646K revenue by designing and shipping the audience infrastructure for a messaging platform launch — including a 45-diff unified ingestion system achieving 85% pipeline adoption and 50% uplift in business uploads 🔧 Recovered 40% of lost mobile device-ID uploads and fixed a 3+ year-old multi-PII hashing defect (10% match rate lift) — both validated via controlled A/B experiments before global rollout ⚙️ Designed a percentage-based phased rollout framework for safely deploying audience enrichment configs globally — full lifecycle ownership from schema design through production, now the team's standard deployment mechanism 🏗️ Led a major platform decoupling (M1→M2 architecture), enabling unified subscriber pools per ad account — resulting in 26.4M new subscribers, 200M+ cumulative additions, and ~90% of accounts migrated without any advertiser disruption
Contributed in building new-gen security-first Identity and Access Management solution for Uber.
• Worked to broaden LinkedIn’s advertising offerings from conventional Mobile and Web devices to Connected TV (CTV) platforms. • Led the expansion of advertiser trust features and brand safety signals for CTV. Enhanced support for custom blocklists and allowlists, integrating seamlessly with 42Matters for pre-bid safety signals. • Designed and implemented a backend pipeline to ingest text files and do regex processing on them to store them in our databases serving. • Achieved outstanding outcomes, reaching 100 million unique devices monthly, with over 99 percent interactions from US-based professionals. Solidified LinkedIn’s CTV presence, highlighting the effectiveness of brand safety measures. • Orchestrated end-to-end integration, connecting to third party servers. Designed a robust system to ingest real-time data, informing bid request decisions. • Achieved remarkable success, contributing to a $600 million revenue growth over two quarters. Highlighted the effectiveness of pre-bid solutions, ensuring a more brand-safe environment for Linkedin advertisers.
Worked with Ads Trust team.