Anoop Singh

Engineered production systems in a for ad tech platform optimizing advertiser outcomes (installs, actions) through CTV campaigns. Spanning real-time bidding infrastructure, ML-facing control systems and large-scale data pipelines.

• Drove campaign performance improvements by conducting experiments on impression frequency, lifetime impression count, audience segmentation and changing bid ranker achieving 15% overall performance gain.
• Developed real-time bid manipulation system to dynamically adjust bids based on given performance signals, which can shift upto 90-95% of spend to top inventory segments.
• Designed margin-aware pacing system to adjust pacing against target KPIs set by customer based on surplus/deficit it has accumulated so far.
• Mentored an intern on a bid frequency control system, proposing and guiding the design of a soft capper mechanism that applies penalty based bid reduction as a more nuanced alternative to hard blocking.
• Reduced infrastructure costs by 45% (~$350k annually) through value-based traffic throttling, filtering low-value bid requests without compromising performance. Resolved critical issue in Bigtable by redesigning key structures, decreasing read timeouts from ~12% to 2.5%, enhancing ML feature reliability.
• Built comprehensive reporting infrastructure that included data pipelines, storage, APIs and third-party integrations for large-scale data exports. Implemented end-to-end solutions for campaign reach computation and external reporting,.

• Enhanced display infrastructure by removing AWS access keys and implementing IAM roles. Added support for Azure Blob and GCP for data uploads and CDNs , enabling deployment in non-AWS environments
• Assisted in eliminating identified security threats for the application and played a key role in establishing deployment jobs, services and surrounding automations.

• Managed a team of 2 backend developers while concurrently mentoring 2 interns, ensuring seamless execution of daily tasks aligned with the product vision
• Created Followers Report that clusters top-followed profiles of given profile list havving up to 3 million, and establishing edges between clusters through common followers.
• Implemented a system to detect profile impersonation among well-known celebrities and brands within a pool of approximately 1.5 billion Twitter profiles.

• Created a system to detect duplications within blogs, news, and web data from various vendors, handling a daily volume of 10 million messages.
• Engineered software for in-depth analysis on Twitter posts, efficiently processing 5 million posts, clustering and connecting similar content for comprehensive insights.

• Contributed to a system for real-time insights across social media, and online platforms, handling ~500 million mentions daily.
• Led the creation of a generic Record Manager model, delivering customer-relevant entities via applied filters.
• Actively worked on-call, fixing OOM, Load, Network, lag, database issues, code failures. coming with an iteration cycle of 5 weeks.