Jul 2018 – Jul 2019 · 1 yr · Bengaluru, Karnataka, India

•  Built Peer Set Model which generates immediate competitors for a given merchant in the market based on its size and industries
•  Developed data products for merchants giving detailed key insights on market share, spend share, opportunities, etc.
•  Built Data Health Check tool for merchants using two years of transaction data to ensure the correctness of the insights
•  Developed statistical models to detect the Merchant Category Code shift anomaly, Pin Routing anomaly and Chase Merchant Services anomaly in merchant’s transaction data and have patents in process

May 2017 – Jul 2017 · 2 mos · Gurgaon, India

• Apparel sales prediction improvement using Deep Learning
• Employed Transfer Learning to improve apparel sales prediction accuracy with extracted image features
• Applied data augmentation on apparel images categorised into six classes on the basis of visual appearance in python
• Employed Convolutional Neural Networks and Linear Discriminant Analysis to extract and reduce dimensions of features
• Achieved an increment of 7% in co-efficient of determination for GBM model with image features than the one without it

May 2016 – Jul 2016 · 2 mos · Bengaluru Area, India

• 1.Debt Instrument: Built a model to determine lending rate to loan call minutes to subscribers as a third party. Implemented Multinomial Logistics Regression to determine lifetime Value, Creditworthiness, churn probability. Applied k-means clustering to do customer segmentation and accordingly get discount rate for each segment
• 2.Hedging Strategy: Devised a strategy for the Debt-instrument to reduce exposure to increase in minute cost Proposed the third party to hold a long position in an American call option written by the network operator Priced the option using the multi-period Binomial Option Pricing model using churn probability to get the volatility
• 3.Chat-Bot: Built a retrieval and semi-generative model of chat-bot for a flight booking application. Simulated data using twitter stream, trained SVM classifier on it, used NLP techniques to generate SQL-Queries. Integrated all to chat interactively with user’s input which is pre-processed using various NLP techniques

Jun 2015 – Jul 2015 · 1 mo · Noida Area, India

• Developed an automated Suppliers’ Risk Monitoring Model on python to rate suppliers on a risk-scale of 0-4 and assign them a risk category by just taking the supplier’s name as input
• The model crawls to various news-sites extracting relevant news-urls of the supplier ,passing it to the scraper to extract the news-articles further cleaning the articles to convert them into a bag of
• words. Using Sentiment Analysis a Naïve Bayes probabilistic Bigram plus Unigram model rates and assigns the category

Jul 2014 – Jul 2017 · 3 yrs · Kanpur Area, India

• Selected as one of the 13 student guides amongst 80 girls to mentor first year undergraduates and help them acclimatize the campus environment and curriculum