As the leader of a Center of Excellence (CoE), I directed a team in building statistical models for the insurance industry, utilizing a variety of technologies including Python, Pandas-Python, Scikit Learn-Python, R, Rev-R, MATLAB, GATE, Stanford-NLTK, and various databases such as MYSQL, MSSQL, PostGres, Vertica, Mongo DB, Hive, and Spark.Some of my notable achievements in this role included:- Developing a risk scoring model that accurately profiled the risk associated with insurance companies, which was then consumed by a re-insurance company.- Creating a prospective customer identification framework to more effectively target potential clients for re-insurance purposes.- Building fraud detection models that allowed insurance companies to save time and money by accurately identifying fraudulent claims.- Creating a web-based framework that served analytics models to various departments within the insurance industry, including marketing, claims, underwriting, and actuarial teams.- Developing a product with key features such as role/project based access, user management, tracking, and web tracking insights, which was integrated directly into the product itself.

As a team leader, I utilized Python, R, and various databases such as MYSQL, Vertica, Hive, and Spark to develop statistical models across multiple verticals. Notable achievements include:- Developing a predictive model that identified emerging technological trends using natural language processing and machine learning techniques. This involved collecting data from various sources, such as Github API and web scraping, and using TF-IDF methodology to identify popular keywords and technologies.- Building a framework using Apache Spark to process large volumes of data and create dashboards for emerging technology trends, as well as integrating statistical models to improve accuracy.- Developing a forecasting model for ad inventory forecasting and processing 20+TB of VOD ad data using Spark.- Leading a team of 6+ people on Market-Mix modeling for a large US retailer.- Building machine learning models to identify factors affecting overall life satisfaction using public data sources such as Twitter and Facebook, survey data, and public bureau data on population demographics and income.- Leading a team in building an end-to-end product for re-insurance business, including model building and visualization dashboards using open-source tools such as Python Flask, D3, Bootstrap, and HTML.

Utilizing a broad array of tools and technologies, including Python, R, MATLAB, GATE, Standford-NLTK, and various databases, I have extensive experience in data analysis and predictive modeling across several domains. Some highlights of my work include:- Leading projects for lead generation, scoring, and recommendation using Twitter data- Developing Poisson/Logistic Regression and Time Series Models to estimate supply forecasting for video-on-demand services- Utilizing Hierarchical/Probabilistic Clustering techniques to identify meaningful business clusters- Applying Natural Language Processing techniques using GATE/Stanford NLTK for Anaphora/Cataphora/Exaphora Resolution, Alias Representation, POS Tagging, topic modeling, and semantic analysis- Conducting social media research and providing insights using public APIs such as FB Graph API, Twitter Rest/Streaming API, Pinterest, Tumblr, Google API, FreeBase, Echonist, and eBayWorking with various classifiers, including Naive Bayes, Random Forest, and Decision Tree- Leading project management and client interactions.

Technologies used include Python, Django, HTML/CSS, JavaScript/jQuery, XML, MySQL, Shell Scripting. Roles & Responsibilities:* Design, architect and implement new features and functionality for the Zebit product (lending business for under-banked people in UK) from scratch. Involved in Zebit database schema design tweaking it constantly to suit requirements. * Developed front-end and back-end functionality to enable the user to complete an application process by providing an user-friendly web interface on desktops, smartphones and tablets.* Worked on optimizations of pages including speed, database query performance and implementing memory caching. * Integrating the Zebit product with merchants and vendors providing Zebit financing by creating API Services. Also worked closely with the marketing and analytics teams to run A/B tests and multivariate tests to improve conversions. * Developed web services to send and receive data from various interacting components like the underwriting engine (which assesses the eligibility of a customer to receive a loan), card services and transaction management systems * Planning, setting up and deploying servers for each of the development, testing, staging and production environments and included them in the release management process.

- Developed a software using MATLAB image processing toolbox which detects and classifies the vehicles from traffic videos by lane wise and type of vehicle(TW, LMV, HMV etc.).- Added the additional functionality for estimating the vehicle speed- Monitored the Chennai traffic (in IITM Vicinity) using this software and benchmarked it against the industry products such as Trazer etc.

B.Tech Project:Title: 3D reconstruction and Meshing of a Heterogeneous Body under the guidance of Dr.U. Saravanan Image processing is an effective tool used to interpolate a complete image from the partial raw data and hence widely used in 3 dimensional reconstructions. This project deals with the application of matrix structured image processing in the study of heterogeneous body and to study various methods of reconstruction and choosing the best suitable method for the complete reconstruction of concrete cylinder. This project is about reconstructing the concrete cylinder in 3D from the sliced 2D images using interpolations and approximations as per the requirements. Next step is writing the code to evaluate the shape defining factors such as Shape, Angularity and Surface Texture for the gravel inside and finding the distribution of these factors by fitting the data distributions and testing the distribution using available standard methods (kstest etc.). Finally reconstruct the gravel inside the sliced pieces and in between the sliced pieces. Meshing up the reconstructed body and studying the variation of different quantities with application of loading using ABAQUS.