Post-Purchase Ads: Architected Golang + GraphQL APIs powering ad delivery across PayPal, Venmo, and

Honey for ~10 million daily users, enabling a new multimillion-dollar revenue stream. Developed additional user opt-out and store allowlisting workflows within these APIs for compliance with GDPR and CCPA regulations.

On-Site Advertisement Engine: Led development on a TypeScript + GraphQL API that processes client text signals and extracts product queries via a fine-tuned Gemini LLM on GCP Vertex AI, with a RAG layer and controlled prompts. Delivered catalog-integrated recommendations blending sponsored listings with organic products, driving new ad revenue and accelerating organic user growth by 15%.

Product Embedding Pipeline: Engineered an embedding pipeline combining reasoning signals, driven by ESCI and CoVE weighted scores, with Honey product catalog SKUs, generating vector embeddings for over 300M products in Elasticsearch to improve semantic search retrieval quality.

Personalized Product Ranking Model: Evolved a winning hackathon prototype into a production-grade

personalization system on Vertex AI. Designed and deployed a bidirectional transformer two-tower model (Python) trained on shopping behavior and product embeddings, re-ranking product SKUs in real time. Outperformed CPM expectations by 108% with personalization, while non-personalized variants underperformed by 76%.

Monorepo Development: Developed Node.JS and Lerna-based monorepos for reused libraries. Setup distinct npm artifactory package publishing procedures. Wrote and deployed concurrent running scripts to constantly monitor and update dependencies in both monorepo structures as well as in consumer repositories. Migrated multiple reused code processes into libraries within the monorepo and built TypeScript support in the webpack processes as well

Backend Search API: Led a quarter-long initiative to build the backend for a new search and indexing feature for Wizarding World. Designed the API using Swagger. Built the API within an AWS API Gateway Contained serverless API using Typescript, with the data modeling and querying being done via Elasticsearch and GraphQL. Completed the development process by building a Github actions-driven CI/CD process using Terraform as well as Cloudwatch-driven monitoring logs and alarms

Full-Stack Engineering: Developed a comprehensive full-stack application using Node.JS, React JS and Express for building, updating and activating store recipes. Built Redux store based state management in the application as well. Incorporated ORM based GCP Spanner interactions within the application. Containerized and deployed the application using a CircleCI based CICD pipeline to Google Kubernetes Engine.

API Development: Built a new API using Node.js and Express for taking in product metrics across various new stores and test automated coupon application crawls. Used Mongo to store the results from the crawler and have it available in an internal dashboard built using React and HTML Bootstrap. This improved the efficiency of new store integration process within PayPal - Honey by 65%. Productionalizing this API, using semantic release, allowed the deprecation of an older API built in Ruby which relied on external vendors and lacked SSO integration.

Manifest V3 migration: Implemented Manifest V3 compliant JS functional code to modify VIM logic within PayPal - Honey and rectify behaviors across the syntax tree for business crucial processes including Find Savings, Product Fetcher and Directly Applied Coupons

Backend Engineering: Developed python scripts in AWS Lambda to auto build Quarterly Business Reviews by procuring data from Snowflake databases and building complex statistical analysis and visualizations. Pushed the project to production within a CICD environment using Jenkins and reduced business review slide generation time for Travel managers by 65%

Data Engineering: Built ETL pipelines in Snowflake using SQL for incremental load summary of Expedia's Login User Data and Travel Manager summary data. Productionalized a Python based sftp script in EC2 to connect the above data to Qualtrics sftp servers. Developed Python based data consumption pipelines of above data with parity checks, PII data encryption and a reduced load time of 93%

Python Scripting: Implemented python functions utilizing boto3, pandas, python-json and cytoolz to create a data warehouse for parsed chat data from Egencia's website and smartphone apps. Created efficient pipelines for this parsed data to be incrementally loaded across S3 and Snowflake for application of Machine Learning algorithms

Teaching Assistant for CSCI 1300. Teaching freshmen algorithmic and object oriented programming in C++. Building assignments and tests for the course. Along with teaching, also grading a group of 85 students.

Voted best TA by student vote for CSCI 1300 in Spring 2020 semester