Alexa Shopping - Product Advisor

Designed, trained, and delivered a Homepage Ranking Model used to re-order and personalize candidate modules on our app. Used a wide feature set aimed to increase user engagement. Worked across the full ML stack from model development to deployment, integrated the model between our PHP and Flask endpoints.

Built out an entirely new radio entity called "Surprise Me". Leveraged our Deep Recos service to use user listening history and various features to make timely and nostalgic music recommendations.

Conducted multiple AB tests aimed to increase activation of new users. Experimented with many recommendation variants on random samples of this user segment.

Rebuilt mood and artist radio song recommendations with a deep learning approach to train song embedding vectors, enabling recos with Approximate Nearest Neighbors (ANNOY) rapid search index.

Authored Airflow offline workflows to serve personalized artist recommendations based on user interactions and impressions.

Implemented video autoplay recommendations for the launch of video entities on our platform. Analyzed and developed a scalable solution for this cold-start problem.

Was a key contributor to the complete rebuild of JioSaavn.com using various open source libraries/frameworks including React, Redux, and Sagas. (Full Stack)

Built the end to end backend flow of our highly successful JioSaavn Pro Rewards service. (Backend)

Built an internal tool to configure and manage all brand partnerships for our embedded player widget platform, as well as an external partner tool to generate and demo new widgets in a self-service fashion

Built an analytical tool that studied the behavior of 11M+ daily active users and evaluated the performance of our content placement engine using the normalized discounted cumulative gain (NDCG) metric. Used Airflow to automate a weekly DAG that uses PySpark to compute the NDCG on weekly experimental changes. This tool will continue to guide/gauge the experiments run on the engine that informs recommendation modules on the mobile app.

Learned how to use and navigate: Amazon Web Services (AWS) Console, particularly using EC2 instances and S3, MongoDB Console to monitor and track the load and performance on databases that support our business needs. Completed Google's Machine Learning Course and learned about various machine learning algorithms that we deploy, as well as how collaborative filtering is heavily used in our product.