Dec 2021 – Present · 4 yrs 4 mos

• Responsible for gathering requirements, ideating and designing solutions to open-ended data engineering problems in collaboration with cross-functional teams, and subsequently developing, deploying and maintaining data pipelines that provide machine learning based solutions.
• Contextual Audiences (Ad Targeting)
• Designed and developed code for the Contextual Feature Store which serves as the starting point for all Contextual applications.
• Refactored code to increase reuse and created reusable modules that can easily be extended to new product offerings.
• Designed, developed and deployed 2 new offerings for this product; Owner of all backend data pipelines.
• Mentoring a recent grad on software development process and providing hands-on training on Spark/Scala, Pig and Oozie since Dec ’21.
• Predictive Audiences (Ad Targeting)
• Worked on a feature for passing user scores from our modeling engine to ad serving for online optimization. This feature required collaboration with multiple teams, understanding the design of components and the required changes, implementing, testing and deploying them. It generated a 2.5% revenue lift, which translates to $7M in additional annual revenue.
• Currently working on enhancing our bucketing algorithm to ensure generation of stable audience tiers for CPA campaigns.

Apr 2019 – Dec 2021 · 2 yrs 8 mos

• Neptune (Ad Targeting)
• Worked on development of redesigned User Profile Store. The previous store consisted of 30 day aggregate profiles and took upwards of 24 hours to build daily, whereas Neptune consists of daily profiles and takes only 1 hour to build.
• Designed and implemented GDPR processing logic for Neptune. The pipeline was designed with performance and resource consumption in mind. It is capable of processing 180 days of user data in under 11 hours.

Mar 2015 – Apr 2019 · 4 yrs 1 mo

• Targeting Sciences - Unified Audiences
• Acquired and cleansed data for Mobile Targeting Machine Learning Platform.
• Suggested use of Mechanical Turk to reduce data sparsity issues and spearheaded implementation.
• Automated NLP extraction and modeling code using Oozie for reducing run time whilst tuning and improving performance.
• Proposed and implemented common CI/CD framework for Oozie jobs across multiple teams based on prior learning experiences.
• $120M annual revenue under influence; Successfully integrated as features into Online Optimizer for Mobile Ad Platform with 2% CPM lift and ∼14M incremental annual revenue.
• Added mobile carrier-based audiences; Projected additional revenue is $35M annual revenue under influence.
• Yahoo AdManager (YAM+)
• Designed and developed Oozie workflow pipelines to massage offline performance data and generate models for CTR prediction.
• Worked on a feature that brought a 500% improvement in runtime.
• Worked on a feature that reduced grid resource consumption by 60% while producing a 10% revenue lift.
• Created Proof-of-Concepts (POCs) on Elasticsearch and Kibana for large-scale data visualization.
• Trained and on-boarded new team members with different levels of experience.