Python Assistant

AGNIJIT DAS-GUPTA

E-mail: ********@****.*** Cell: 415-***-****

EDUCATION

University of Delaware, Newark, DE, USA: M.S., Computer & Information Science (August 2017 – May 2020) University of Delaware, Newark, DE, USA: M.A., Economics (February 2009 – January 2013) University of Delaware, Newark, DE, USA: M.S., Operations Research (August 2007 – December 2008) University of Kalyani, India: B.Tech., Computer Science and Engineering (June 2000 – May 2004) TECHNICAL SKILLS

PROGRAMMING LANGUAGES: Python SQL R

DATA SCIENCE: Pandas NumPy SciPy Scikit-Learn OpenCV Data Visualization (Matplotlib) MACHINE LEARNING: TensorFlow Neural Networks (CNN, RNN, LSTM) Unsupervised Learning (Clustering, PCA) Random Forest K-means Clustering Naïve Bayes

INDUSTRY KNOWLEDGE: Banking Finance Client-Facing Consulting Retail eCommerce Media EXPERIENCE

Instructor: CISC220 Data Structures and Algorithm Analysis, University of Delaware (June 2018 – Present)

• Teach a class of 30 students in the College of Engineering

• Boost student moral, experimenting with presentation styles to increase class participation which is key to success in the course

• Design the course material, administer the lectures, laboratory setting and examinations

• Mentor students to excel in their career objectives, improving efficiency of communication

• Manage the Teaching Assistant for the course to help the students when needed Research Assistant, Video/Image Modeling and Simulations Lab, University of Delaware (Sep 2018 – Dec 2019)

• Improving Deep Neural Network models’ performance using Tensorflow as backend (Python)

• Researching on using GPU acceleration for the Deep Learning models

• Enhance Virtual Reality models by using stereo cameras, calibrating those and capturing images for depth sensing

• Estimating disparity of stereo images of ice-covered waters with low texture information, as a starting point of reconstructing 3D

• Apply the algorithms on Middlebury test data sets to compare against their ground truth disparity values with S-O-T-A results

• Most implementations available on github: https://github.com/VimsLab/Stereo-Ice-Reconstruction

• Publication at CVPR - Color And Thermal Stereo Scenes with Semantic Labels: http://openaccess.thecvf.com/content_CVPRW_2019/papers/Vision%20for%20All%20Seasons%20Bad%20Weather%20and%20Nig httime/Treible_CATS_2_Color_And_Thermal_Stereo_Scenes_with_Semantic_Labels_CVPRW_2019_paper.pdf Practice Head Machine Learning, Velotio Technologies (April 2017 – June 2018)

• Responsible for building the Machine Learning Center of Excellence

• Lead a team of 20+ Data Scientist & Engineers to build predictive machine learning models using R, Python (utilizing Random Forest, Gradient Boosting/AdaBoost, Deep Learning algorithms)

• Slashed cost and generated fresh revenues implementing predictive models such as customer churn, propensity for subscription, and audience engagement based on content

• Identified and executed more than $4MM in cost savings by implementing a new demand forecast process to eliminate production of low performing products that were not revenue drivers.

• Identified key markers in a project where the turn-around time was reduced from 9-12 months to about 9-15 days within contract management using AI and machine learning approaches which helped save many lives by a top pharmaceutical organization.

• Launched the first home-grown analytics platform for providing insights into story-level performance and prediction of certain metrics that drive business decisions

• Setting enterprise data strategy and worked cross-functionally with Product, Marketing, Advertising to integrate and rollout analytics toolsets including MS AzureML

Director Data Science, Amazon (July 2015 – March 2017)

• Improved customer acquisition models (probability/scoring) for the US marketplace using Python and R Studio to 89% (F-score)

• Traveled frequently to the Euro zone to prepare a Data Science team to build acquisition models for Germany and UK marketplaces

• Increased the customer base/memberships by 9%-11% over previous years after implementing the models

• Enabled the production team to reproduce the predictive models in Python (utilizing Gradient Boosting/Deep Learning technique)

• Acquired data using ETL builds from Audible and Amazon data warehouse (EC2/DMART/Redshift)

• Achieved lower runtime of queries/algorithms (by over 70%) for best performance and minimal resource usage

• Enhanced customer retention strategies using the results from the models, increasing revenues by over 16MM

• Improved customer satisfaction manifold over the months by collaborating with Customer Service (CS) to understand the business process and built models (likelihood of a person to Cancel/Convert) to support CS Representatives that expedited call-handling capabilities

• Collaborated with current research on Machine Learning (ML) in Amazon Conferences and workshops primarily using Python

(Deep Neural Network – DNN, with Keras framework using Tensorflow as backend) Data Scientist Consultant, CableVision Media Sales (Nov 2014 – Feb 2015)

• Created customer segmentation and clustering models using Python, SPSS Modeler & Netezza Database

• Researched and upgraded models using Survival Analysis and Python(Tensorflow DNN) to detect and predict customer churn

• Optimized customer retention strategies using the results from churn models, reduced by about 21%

• Implemented Machine Learning algorithms to support decision making with Treenet & Python Sr. Database Marketing Modeler, Conde Nast (Mar 2013 – Nov 2014)

• Boosted sales volumes for each of the publications (quarterly) by implementing state-of-the-art cross-sell, list rental, expire, agent expire and other models using current machine learning research

• Designed customer acquisition models for in-house magazines and for a few outside retail vendors as a part of our partnership program, hence increasing the customer base and total revenues for each quarter by ~ $11MM

• Surpassed previous mailing campaign response rates by applying efficient analytics that increased subscriptions by 19%-23% overall

(in 18 categories combined, from previous years)

• Performed critical in-depth analyses for annual mailing campaigns and met with leadership teams to enhance marketing strategies for monthly and seasonal mailings

• Recruited top technical talent to maintain the models

• Mentored them on advanced methods of analytical marketing techniques using Python, R Studio, SAS and other tools

• Improved the efficiency of targeting models, both at algorithm runtime and result (Net Response Rate) levels AVP, Consumer Products Strategy Manager for Forecasting

(Univ. of Delaware CPT work auth ending Summer 2011), Bank of America (Jan 2011 – Aug 2011)

• Directed a team of Senior Analysts responsible for short and long-term forecasting of recovery collection dollars for Domestic Card Products, and for performing strategic analysis to improve portfolio risk, profitability forecasting and operational performance for consumer credit cards

• Established use of forecast models for Consumer Card and Consumer Lending products and presented analysis of variance drivers to senior management. This model was applied to other LOBs after current implementation using R Studio

• Achieved an accuracy rate (>85% for 12 months) in the forecasting models that haven’t been possible earlier

• Improved the analysis focusing on criteria margins, model score use, forecasting, portfolio trends, and demographics

• Developed multiple forecast scenarios based on statistical analysis of economic variables and potential forecast drivers

• Mentored team members on Statistical Analytical processes and techniques using R Studio, SAS/ETS and SQL, used for developing, optimizing and improving the models

• Acted as a single point of contact for all data management, Python and R program implementation needs for the Risk Organization

• Assisted in the Vendor Goal Setting process for the Sales Strategies Team to maximize recovery dollars by designing an optimized vendor allocation system which surpassed previous numbers Developer Analyst, Intern, Wilmington Trust Company (July 2010 - Sep 2010)

• Streamlined the Customer Relationship Management process in the Wealth Advisory Services

• Enhanced sales by analyzing data using R Studio and SQL for cross selling products

• Instrumented dividing the customer base into segments – Affluent, Mass Affluent, etc. primarily using machine learning in R

• Presented the senior management with the key findings and proposals for improving the customer experience, depending on their feedback by sending more targeted products, preferences and participation in events sponsored by the Company

• Generated forecast models for the participation rate of customers in events and in adopting new products offered Graduate Analyst, University of Delaware (Univ. of Delaware Graduate Assistant) (Sep 2007 - Dec 2009)

• Development of econometric and machine learning models - 1. Built an econometric model for the state of Delaware using two-stage least square technique to estimate man-hour, wage and employment equations, and then putting these for twelve different industry sectors into a simulation model. This model can be used for forecasting purpose and policy changes

2. Estimated the effect of lot size on asking price of houses in Delaware using regression model in R Studio and SAS 3. Developed a model to inexpensively measure Body Fat percentage using easily measurable variables like weight, height, age, abdomen circumference and putting them in a regression model using R Studio 4.Established a model for assessing wage increase for increasing participation rate in the labor force and increasing the hours worked for Registered Nurses in the US using selectivity corrected equations in SAS and LIMDEP. This is an economic analysis of the current RN population, revealing their behavior towards wage changes, depending on their marital status, family income, education, region and other factors, corrected for selectivity bias

5.Published the work above, in a thesis paper named: FACTORS AFFECTING THE LABOR SUPPLY FOR REGISTERED NURSES IN THE US

Assistant Systems Engineer, Tata Consultancy Services (Sep 2006 - July 2007)

• Implemented the SAP SRM project for Tata Iron and Steel Limited as team-lead for the implementation team with nine analysts.

• Trained associates on the new system and made everyone feel more confident about usage, as they were changing from the legacy system to computerized systems in a very short period of time ERP Consultant, Yugine Technologies Ltd. (June 2005 - August 2006)

• Managed a group of five software analysts to implement the SAP MM module and resolved any issues that came up in the process of RFP, PO, Goods Receipt transactions

• Trained Quantum Foods Inc. associates on using the SAP MM system to automate the business and went on to support

• Improved the ongoing support system for Quantum Foods Inc. by introducing CMM Level 5 approaches Software Engineer, ITC Infotech India Ltd. (June 2004- May 2005)

• Tested the Materials Management and Business/DATA Warehouse modules of SAP for DHL Scottsdale

• Contributed ideas in the support system for SAP DHL project

• Diagnosed the day to day issues in MM module and reported problems using MercuryQualityCenter, SQL, and Data Mining CERTIFICATIONS

SAS Enterprise Miner

Scuba Diving (PADI)

RELEVANT EDUCATIONAL COURSEWORK AT THE UNIVERSITY OF DELAWARE Artificial Intelligence, Computer Vision, Database Management Systems using SAS, Logistic Regression, Probability Theory, Data Mining, Big Data Analytics (PySpark), Game Theory, Analyses of Algorithms, Statistical Research Methods, Discrete Event System Simulation, Optimization Models using Spreadsheet (Excel – Premium Solver), Advanced Econometrics, Micro & Macro Economics, Time Series Analysis, Mathematics for Economists, Economics of Regulation & Prices, Investment Analysis & Portfolio Management, Industrial Organization

Contact this candidate