Yash Parekh Email & Phone Number

San Francisco, Ca, Us

Seattle, Wa, Us

◦ Amazon Rufus (ShopGPT): Worked on new Autocomplete and adaptive CX Tier 1 services (180K TPS) and scaled them from 1000 to 100 million customers launching the services to multiple marketplaces. Delivered a realtime system leveraging billions of customer interactions daily from Amazon’s Conversational AI for corpus expansion of the Rufus search autocomplete suggestions.◦ CMAB and MLOps for adaptive CX: Developed a continuous Multi-Arm Bandit (MAB) using Amazon’s shopping data to predict CX configurations based on customer engagement. Leveraged SageMaker and EMR to architect the SageMaker pipelines to support big data to automate ML model training and deployment.◦ Search Token Generation: Spearheaded an initiative to scale the Search Question Answer Widget for Shopping Actions (providing related search items for the Alexa FAQs) from 4K keywords to a million to achieve a 100 million MAU (monthly active users) target using generative machine learning techniques.◦ Performance Evaluation Pipeline: Developed a cloud-native solution for the Search Question Answer Widget to compute and report the financial performance metrics of search keywords.

New York, Ny, Us

Centralized Data Extraction library (msci.data) – Developed an enterprise-wide data extraction library (extensively being used by 1000+ employees and supports more than 15,000 datapoints) used for performing financial analysis and publishing reports to clients; providing fast, reliable and convenient way to access data from various data sources (internal as well as external) on MSCI’s Data Science Platform. Tech Stack: Python, HDF5, MongoDB, Redis, HDFS, Cassandra, Redis, Flask, Memcached, Oracle SQL, MySQL, Apache Ignite, Apache Spark, Jenkins, Docker, Sphinx, Git, Talend Section Extraction Service (msci.sections) – Developed Section Extraction API which breaks company filings (US, Canada and China) down to modules of interest translates the language of the text and delivers data through a RESTful API. The API improved the accuracy of the business-critical machine learning models by over 15% which were earlier using vendor data from Wordscope and Factset API and the search UI helped improve the efficiency of Financial Analysts multifold providing coverage of more than 95% of the filings becoming a primary data source of business description for Thematic Index based products. Tech Stack: Python, Fast API, jQuery, JavaScript, HTML, CSS, MongoDB, Oracle SQL, Elastic Search, Azure, Sphinx, GitText mining on Lexis Nexis news data – Led an initiative to improve coverage of dividend data from the news sources having an accuracy of more than 95% in deriving dividend information.Tech Stack: Python, Fast API, Azure, Apache Spark, BERT, TF-IDF, Word2Vec, genism, SQuADData Science Platform –Involved in development, deployment, optimization, providing trainings and maintenance of the centralized JupyterHub platform at MSCI and the common libraries used by 1000+ employees for building analytics and reporting solutions.Tech Stack: Kubernetes, Docker, JupyterHub, Python, Shell, Putty, JIRA, Azure

New York, Ny, Us

Data QA framework (msci.qc) – Developed a QA framework which could apply various data transformations, perform EDA and apply machine learning and statistical models to a data feed and also generate reports based on it. The framework is used to perform quality check for over more than 100 datapoints in production.Tech stack: Python, Java, Git, Shell, Oracle, Flask, Docker, Flask API, Java, Ext JS, Spring, JIRA, Putty, SVN, Apache TomcatData Quality Check – Contributed to the enhancement of the data quality by developing QA models to detect and report anomalies in the Fundamental data, ESG data and Investment Broker Estimates datapoints using various supervised, unsupervised machine learning and statistical techniques. Tech Stack: Python, Sklearn, Numpy, Pandas, Matplotlib, Altair, Keras, XgboostMSCI’s Automated Thematic Index Generation System – Led an innovative development effort of MSCI’s Natural Language Processing based thematic index products (estimated worth $1 billion) in collaboration with Index Research team. Worked on web crawling, topic modelling, sentence embedding, neural network and deep learning techniques to provide a probabilistic classification of the theme of an equity based on its business description. Our team received appreciation and recognition for the accomplishment of the project by directors at various level and the solution is currently being delivered to clients through Thematic Index products. Tech Stack: Python, NLTK, Word2Vec, Keras, Pytorch, Docker, MongoDB, Web Scraping, FastText, Gensim, Pandas, Tensorflow, SeleniumSematic based Voting rights classification model – Built a semantic model to determine whether a security has voting rights for its shareholder class or not using NLP, text mining and pattern recognition techniques.Tech Stack: R, R Studio, Shiny app

Mumbai, Maharashtra, In

Independently involved in the end-end development of the following modules – 1. Chat module - Text communicator for two or a more users 2. Notifications - Event based application alerts to the user 3. Data Visualizations - Report data driven visualizations 4. Selenium based web crawler for data extraction from the webThe application is being used my more than 15 hospitals including Fortis Healthcare.Tech Stack: MongoDB, NodeJS, ExpressJS, AngularJS, React Native, HTML, CSS, JavaScript, jQuery, Selenium, Bootstrap, AWS S3, PhantomJS

Mumbai, Maharashtra, In

Implemented an inventory management SAP mobile application system. The application allowed inventory and supply chain teams to manage raw materials for efficient manufacturing of finished products.Tech Stack: SAPUI5, HTML, CSS, JavaScript

Master'S Degree, Computer Science And Engineering

Bachelor Of Engineering (B.E.), Computer Software Engineering

Hsc, Science

Icse

Education record