Brooklyn, New York, United States
Leading AI-driven data solutions at Meta, my focus lies in enhancing infrastructure with a robust foundation in ranking and AI. At Google, I honed skills in data ingestion and local indexing, ensuring food query accuracy through machine learning validations. My work contributed to seamless integration of dietary preferences into search functionalities, benefiting users globally. With a solid educational foundation from the University of Virginia, majoring in Mathematics and Computer Science with a minor in Philosophy, I bring a unique perspective to technology and problem-solving. Certifications in Blockchain, Big Data, and Docker complement my skills in BigQuery and BigTable, augmenting my contributions to Meta's innovative endeavors. Made with AI, so if it doesn’t sound like me, that’s why.
Designed and iterated on ML feature pipelines incorporating external search signals and offsite pixel ID data to improve ad ranking model performance, evaluating model behavior across experimental feature branches. Integrated Claude Code into feature pipeline development workflows, using AI-assisted code generation to automate boilerplate, streamline codebase cleanup, and accelerate feature importance experiments — building a feedback loop between model evaluation results and pipeline iteration. Adopted Random Response Differential Privacy (RR-DP) in ad ranking models through rigorous experimentation, measuring statistical significance of privacy-utility tradeoffs. Built graph-based and embedding-based cohorting experiments to improve model accuracy and user privacy, analyzing performance regressions and edge cases across cohort methodologies. Engineered event-based features and supporting data infrastructure for ads distribution, owning the full pipeline from dataset preparation to model evaluation. Researched and implemented denoising techniques for user signals in event-based features, evaluating impact on feature utility and downstream ranking model performance. Serve as model owner for ad ranking model architecture and feature refreshes, leading evaluation of proposal combinations to assess performance tradeoffs and drive model improvement decisions. Developed feature pipeline infrastructure to enable rapid experimental iteration of event-based features via feature injection, reducing the cycle time between hypothesis and model evaluation.
Developed infrastructure and processes to ingest price information from partner-, merchant-, and user-provided information of offered food at locations. Developed local indexing pipeline for food-at-location information for higher local query data accuracy. Designed and implemented validation infrastructure for ingested food data to handle scalable changes to the corpus. Designed and developed data ingestion platform and aggregation for dietary attributes such as vegan, vegetarian, and taste profiles using machine learning predictions on user and merchant submitted data. Coordinated efforts across teams in the New York and New Delhi office to design and develop aggregation scoring for trusted user contributions with respect to food attributes and data quality. Interviewed candidates for software engineering positions at Google, country-wide.
Created and maintained a partner-facing platform to edit and update copyright information for video assets. Claimed ownership and developed team’s flagship prediction service API for collecting and storing feature information for video and caption assets as a preprocessing mechanism for our content descriptor ML models. Coordinated and represented the team as an away team engineer in creating an automated backfill service for assets to be re-categorized through updated models to maintain high classification accuracy. Mentored a new hire and aided in onboarding to team-owned services and codebase. Received accelerated coursework with respect to computer vision techniques and model engineering methods. Designed and engineered a customizable pipeline library for aggregating and preprocessing video embeddings for repeat predictions on trained models.
Remodeled part of the company’s central web application to gather and display metrics on clinicians’ schedules using TypeScript, JavaScript, C# , and React.js-based framework, and ASP.NET. Maintained critical functions of the scheduling and medical access division of the product in a time-critical environment, involving numerous other bug fixes on client, server, and database. Programmed against screen readers and other assistive technologies to ensure digital accessibility.
Developed a Kubernetes server resource tracking dashboard using python, JavaScript, React.js, HTML, CSS, and docker, allowing interactive viewing of historical trends and current usage of IBM's proprietary AI and Cloud. Gained better understanding of corporate agile practices. Learned fundamentals of blockchain and dockerizing.