• Develop automated systems to detect compliance issues and fraud in Amazon’s Associates program using technologies such as AWS Step Functions, Lambda, SQS, DynamoDB, etc. to execute compliance actions at scale.

• Maintain offline data pipelines with AWS Glue and Redshift, enabling investigators and automated systems to process and analyze large datasets.

• Led service migration to new AWS region and improved operational excellence.

• Mentored intern to design and prototype Generative AI–based services for content policy violation detection.

• Contribute to design reviews, code reviews, and on-call operations to maintain service reliability and availability.

• Built and customized ServiceNow apps (PSDS, CSM) for a PoC supporting Washington State’s court-ordered mental-health rehabilitation process.

• Designed data flows and tailored UI with JavaScript to streamline patient case management; emphasized performance and data integrity.

• Integrated generative-AI APIs to prototype diagnosis support and patient-notes analysis within ServiceNow.

• Designed and deployed big data pipelines using Apache Spark on Google Cloud Dataproc to support large-scale ETL operations, processing millions of corporate records daily, incorporating it into Sayari Graph

• Built directed acyclic graphs (DAGs) in Apache Airflow for end-to-end orchestration of ETL workflows, improving operational reliability.

• Leveraged Google BigQuery for advanced querying, ensuring scalability across billions of data points.

• Developed Scrapy-based web crawlers on Google Kubernetes Engine (GKE) to automate corporate data extraction from public web sources.

Team - Traffic and Marketing Tech

• Designed and developed a complete API service for retrieving eligible Amazon programs to creators to promote on their Amazon storefronts,. Deployed service to handle 300+ transactions per second.

• Implemented full CI/CD (Infrastructure as Code) with unit, load, stress, and canary testing.

• Employed a suite of AWS services like ECS with Fargate, Lambda, DynamoDB, CloudWatch, etc., for efficient scaling and monitoring.

Center on Knowledge Graphs

• Transformed a NoSQL database into a Neo4J-compatible knowledge graph with 17B edges and 400M nodes. This graph built using various graph database techniques is used to study author success, citation & collaboration patterns and citation intent.

• Implemented a pipeline using ElasticSearch and lexical similarity matching to link 5M records for data completeness.

• Utilized high-performance and parallel computing for graph analysis techniques like community detection and snowball sampling.

• Produced 2 workshop papers on topics of author success and citation intent classification.

• Analyze data and matched researcher data to voter records on names using EM algorithm.

• Mine time-varying researcher affiliation data from publications to uncover work history and geocode work addresses via Google Maps

API, to group data to metro areas, reducing record matching scope to a smaller area.

• Enhanced organization name matching, resulting in an increased match rates to create high-quality ground truth data.