Chavi Gupta Email & Phone Number

Redmond, Washington, Us

Redmond, Washington, Us

Redmond, Washington, Us

Generating Differentially Private Synthetic Datasets using Normalising Flows

Redmond, Washington, Us

I worked on building a real-time data analytics platform for Azure Traffic Manager which enables customers to view traffic for globally distributed endpoints. Improved customer experience by reducing data latencies from 28 hours to 15 mins (~99% reduction). Reduced latencies enabled customers to view near real-time traffic. By adding time dimension to the traffic data users can perform time-based analysis for scaling their services and resources. Real-time traffic also helps in quicker fault detection and resolution.

New York, Ny, Us

Tools used - Java, Springboot, Oracle 12c, Bitbucket, Git, Jules CI/CD tool, Agile Methodolgies- At JPMC I worked on an end-to-end cross-site (Chicago, Tampa and India) application that reported and reconciled receivables (payments) coming in from various remitters and modes of payments for various businesses across the USA and other countries. The application also provides a feature for clients to directly scan checks which are digitalized, balanced, validated and then sent to the clearing house. Eventually it is reconciled for further processing. - I worked as a developer to build microservices for scanning, validating, balancing and submitting checks for processing. I also built microservices for reporting payment summary centered on different variables for different clients.- Won 1st place in the annual FlowMonster contest in 2018. The contest involved building an automated trading system and algorithm, to compete in intense client-dealer markets to gain maximum client market share and profit. - I mentored two interns during their 6 month internship at JP Morgan, who later joined the team as full-time employees.

Berlin, Berlin, De

Tools used - Java, Linux, Play framework, PSQL, Druid Db, Ignite Cache, Git, Jenkins CI/CD tool, PythonZeotap is a Germany Based Startup in the Advertisment Technology(AdTech) business. Zeotap is a Data Management Platform, offering end to end data solutions to marketers, publishers and data owners. Zeotap Audience Insights is among the major data products owned by Zeotap. It is a platform for all marketers and publishers to get audience demographics and data for data analytics, for targeted and thus effective advertising. The analysis of audience data gives rich insights to businesses and advertisers about their products and is used to help in different phases of businesses like product design, business modeling and targeted advertising.Zeotap pairs up with 1st, 2nd and 3rd party data owners to get high quality data. The massive amount of data is cleaned and partitioned into different segments through big data techniques and offered to publishers and advertisers through the Zeotap Audience Targeting platform.During my internship at Zeotap I had the opportunity to work with both of Zeotap's major products. I gained insight on working with high volume data and systems while maintaining security and complying to GDPR. I also got the opportunity to own the backend system for Zeotap Audience Insights v2.0. Besides having better data visualizations than v1.0, the application performed 10 times faster at data retrieval and pixel creations due to the use of Ignite Cache.I also helped to apply privacy controls and risk policies to comply with the General Data Protection Regulations (GDPR).

Redmond, Washington, Us

Tools used - R, C#, Excel, Azure ML StudioAt the Microsoft Technology Center, I was focused on solving the Candidate Hiring Problem using Machine Learning. Every large organization Since Microsoft receives thousands of resumes from applicants everyday. Shortlisting right profiles is an interesting Big Data problem that we tried to solve using Machine learning concepts.We made a resume parser to pick up applicant skills and specializations and built a word cloud to visualize the most common skill set vs the most sought after skill set in different industries. We also used a data set of over 30 thousand applicants. I processed the data set through many stages of data cleanup in R and performed Exploratory Data Analysis, feature reduction and PCA. I then used unbiased partitioning techniques to split the data set for training and testing. We trained multiple models on the training sets to predict whether an applicant is likely to be hired. We ran the predictor on the test sets to gain 94% accuracy on the predictions. The predictor was thus an effective tool for filtering out candidates less likely to be selected based on work experience, skill set, industry role, location preferences, etc. This was my first step into machine learning, and I got the opportunity to learn the different stages of building an ML system including data cleaning, analysis and model training and testing. The mathematics behind the models excited me and the end rewards were satisfying. Along with machine learning, I had the opportunity to learn and work with new languages (R and C#) and the Azure ML studio.

Chennai, Tamil Nadu, In

Tools used- Node.js, HTML, CSSRisk management is a key aspect for Banks to determine Credit worthiness of any customer.Indian Bank is a large Public sector bank that have defined risk assessment model and parameters that they choose to use for such assessment.I implemented a prototype of a web application for this risk assessment module, including the frontend which accepts various input parameters and the backend which determines the credit worthiness of prospective customers.

Master'S Degree, Data Science

Bachelor Of Engineering - Be, Computer Science

Msc, Mathematics