Nov 2021 – Nov 2024 · 3 yrs · Mumbai Metropolitan Region · On-site

• Data Mining and NLP in Credit Underwriting using BERT for SMS Data: Utilize advanced NLP techniques to extract valuable insights from SMS data to improve the AUC & K-S of credit underwriting decisions
• Develop a fine-tuned BERT model using the TensorFlow framework to classify SMS messages as either useful or non-useful for credit underwriting decision-making
• Batch Inference of DistilBERT Model using TensorFlow Serving in AWS SageMaker TFX
• Developed a comprehensive TFX pipeline encompassing data preprocessing, DistilBERT model training, and validation components, ensuring a structured workflow hosted on AWS SageMaker
• Enable more precise identification of creditworthy and high-risk applicants, reducing default rates
• Application Scorecard: Build an application scorecard to replace pre-existing framework to acquire new customers for personal loan. Boosted AUC (Area Under the Curve) by 8%
• Developed variables using bureau data and SMS-based alternate data to create a more comprehensive view of customer creditworthiness and enhance predictive power of the model
• Leveraged reject inferencing methods to reduce bias and improve fairness of credit scoring
• Conducted swap-in and swap-out analysis to identify an efficient deployment strategy
• Behavioral Scorecard: Built Credit Ensemble Model for Customer Behavior Analysis Enabling Upsell and Cross-Sell Strategies
• Formulated an ensemble model to comprehensively evaluate customer behavior, drawing insights from DPD (Days Past Due), Repeatability, Bureau (CIBIL/Equifax), and DPD Agnostic models, culminating in an impressive 4% portfolio risk for 90+ NPA (Non-Performing Asset)
• Deployed the ensemble model in a batch processing context, utilizing Snowpark for inference and Airflow (DAGs), enabling efficient and automated execution

Nov 2021 – Nov 2024 · 3 yrs · Mumbai, Maharashtra, India · On-site

Aug 2021 – Nov 2021 · 3 mos · Singapore

• Lead brand growth workstream
• Developed custom integration platform for Shopee & Lazada
• Designed data lake architecture to manage acquired brand data
• Time series Modelling: Investment Portfolio Optimization, analyzing patterns/collective trends in
• customer shopping /products/categories based on geography and different months/seasons.
• Competitor Analysis Model: ML model to find all the similar products selling across different
• platforms based on image, product description & keyword comparison, and deploying bots to keep
• track of these products, their pricing and demand to give optimal pricing structure and
• recommendations.
• Developed predictive models for customer/brand acquisition, created pipeline that analyses
• messages/reviews/chats related to specific products or brands to gain insight into product sentiment
• for use in product/product category research.

Jan 2021 – Aug 2021 · 7 mos · Singapore

• Member of the core team
• Developed scalable systems to crawl and scrape data from E-commerce website (APAC region) using Python| Scrapy| Request| Selenium
• Built NLP model to perform sentiment analysis on customers' review
• Dockerized the scraping scripts and deployed them on AWS ECS Fargate
• Established data lake infrastructure on AWS S3| Glue| Athena
• Developed Dashboard to identify potential e-commerce businesses to acquire using AWS QuickSight

Sep 2020 – Dec 2020 · 3 mos · India

• Data Science Architect having experience in Machine Learning, Deep Learning, Artificial intelligence, Reporting, Predictive Modelling, Data scraping, and Dashboards.
• Tech stack: Python (Numpy, Pandas), R, Tableau, QuickSight, Power BI, Tensorflow, Keras, NLP, Scrapy, PySpark, Excel, MongoDB, S3, AWS (EC2, ECS, EMR)

May 2020 – Aug 2020 · 3 mos · Boston, Massachusetts, United States

• Working for a project in creating a matchmaking website that connects front line researchers with people who can help them for the researchers at Broad Institute of MIT and Harvard who are fighting the COVID-19 crisis
• Created interactive dashboards using Dash-Plotly and Tableau for COVID-19 analysis for the website
• Collaborating with scientists from NASA-JPL to query knowledge graph in RDF format using SPARQL
• Managing a team of 15 student volunteers contributing to this project and coordinating their tasks

Aug 2019 – Nov 2019 · 3 mos · Boston, Massachusetts, United States

Mar 2017 – Dec 2017 · 9 mos · Mumbai Area, India

• Performed feature engineering, feature selection, Data visualization with Tableau and utilized Classification algorithm to forecast potential clients and influencers; enhanced overall performance of the system which improved KPI by 37%
• Anayzed valuable statistical insights on potential social media client Profile
• Built a Decision tree classifier model for the business to determine the potential delinquent client accounts on social media activity, thus giving accuracy of 78%
• Formulated Tableau dashboards and developed SQL queries for reports based on SQL Server Reporting platform
• Deployed model using AirFlow data pipeline on AWS EC2 instance
• Containerized the application and data pipeline using Docker

Aug 2016 – Sep 2017 · 1 yr 1 mo · Mumbai Area, India

• Worked with Brokerage Trading Services (BTS) under Value Added Services unit on the development of Online System for Financial Marketplace
• Built a Node.js architecture using Express framework to develop a generic E-commerce website
• Developed front-end for web site maps, application models, image templates, and page templates on BlurAdmin framework using Javascript, JQuery, Angular.js, Sass, Bootstrap, and Gulp
• Developed scalable databases that support web applications using MongoDB
• PoC for deploying the website on Amazon Web Service using EC2 and Elastic Block Storage