About

Mansi P Shah is a dedicated Software Development Engineer at Amazon Web Services (AWS) with a passion for learning, hard work, and analytical thinking. She is an optimistic team player who aspires to contribute to society through her work. Currently a Research Intern at ISRO, Mansi has been working on a Hybrid Recommender System for Satellite Data Products using Python, Statistics, and Machine Learning concepts. Her work involves disseminating around 200 types of data products to researchers and academicians through online data ordering systems. She has found similarities between products by combining similarity between features of products, example Satellite, Payload, Parameter, Domain, Level of Processing, etc. for Content-Based Filtering. She has also found similarities between products by comparing the ordering pattern of users for Collaborative filtering. Her work includes predicting the rating given by the user to a product considering ratings given by the user to similar products. She has hybridized methods to get optimum results using weighted average, switching, cascading, feature augmentation. Her work has resulted in a Mean Absolute Error of 0.5 for ratings in the range of 0-5, with other evaluation metrics like Diversity, Coverage, Recall, and Precision. She has also read research papers to learn about and implement new methods. Mansi has previously worked as a Mentor at Robotics and Coding Academy, University of Southern California, where she taught 4th and 5th grade students to code in C language and build robots through interactive sessions and hands-on experience. She has also worked as an SDE Intern at Amazon, where she created a platform to run simulations of changes to algorithms in the warehouse to evaluate how these changes will affect performance. She filtered current inventory picture data, having around 275 million items by warehouse, and copied it to another DynamoDB by making Data Pipeline, EMR, Hive Copy Activity, VPC, etc., using Java SDK. She saved inventory movement messages from the current to the end date in DynamoDB, after polling SQS message queue that has subscribed to SNS topic generating messages, deserializing message and filtering messages by warehouse. To run the simulation, messages can be replayed for each algorithm. She calculated metrics zero in prime and the number of movements from reserve to prime. She coordinated with the manager, mentor, and buddy for understanding requirements, designing the system, and selecting approaches for implementation. She interacted with the team for code review and presented the project to 30 people of the team. M