Feb 2019 – Jan 2020 · 11 mos · Bangalore · On-site

• Peril detection Framework – Python, TensorFlow, CaffeJS
• Developed a Framework to verify the perils in a claim using the claim assessor’s images
• Defined a use-case for future Drone Usage and RPA in Claim Processing
• Used VGG-16 weights in Keras to build initial model
• Added 5 more convolution layers on top of VGG-16 to add more specificity towards assessor’s images
• Implemented the model in production on CaffeJS
• Health Insurance Claim Invoice Classification – Python
• Developed an unsupervised algorithm to classify the claim line-item level invoices in meaningful clusters such as medicine, implant, Surgeon’s fee, Injection, Pathological Test etc.
• Crawled multiple websites to generate vocabulary for TF-IDF generation
• Weighed the TF-IDF with glove 300 dimensional vectors
• Developed a K-means model to get required clusters
• Final clusters were used in pricing exercise – Frequency-severity modeling

Jun 2014 – Feb 2019 · 4 yrs 8 mos · Pune Area, India · On-site

• Propensity and Churn Modeling for broker activation and deactivation – Python, R
• Defined the problem of broker activation and deactivation as propensity and churn modeling problems
• Data Cleaning and feature creation using Python and R
• Created gradient boosting models with an AUC of the 0.79 and 0.74 for the above problems
• Helped put the machine learning solutions into production
• Risk pricing using Machine Learning techniques– Motor & Health – R, SAS
• Performed risk pricing and modeling for motor book of AXA Mexico & Germany (Personal Lines Intermediary)
• Modeled various motor perils (AD, TPD, TPI) using R and SAS
• Zoning classification – Motor & Health – Python, Classifier
• Performed zoning classification for different motor perils (like collision, theft, glass) and health business
• Performed Smoothing, Spatial correction, Credibility weighting
• Analyzed using Steepest curve test/ HOS test, QQ plot, Maps
• Competitive Market Analysis (Market Observation) – Python
• Reverse engineering and analyzing the pricing structure of different competitors for Germany market
• Impact analysis of proposed change on Key Performance Indicators - loss ratios/ volumes etc.
• Automated Web-crawling & Dashboard creation – Python, VBA
• Built automated web-crawling scripts in python to download prices automatically from different websites/aggregators
• This is used to understand pricing systems, test live rates and compare different products on pricing

Jun 2011 – May 2014 · 2 yrs 11 mos · Noida, Uttar Pradesh, India · On-site

• Predictive modeling - Conversion/Retention Models - SAS
• Built Predictive models to calculate price sensitivity across different channels (Phone/Web) through which policy is sold
• Rate reviews and Impact analysis – SAS, R
• Rate-reviews and impact analysis based on different factors (like No Claim Discount, TV Region) for different products of General Insurance
• Reviews included detailed study of key metrics such as claim frequencies, loss ratio, burn cost, profitability, conversion etc.
• Report automation - VBA
• Converted the Excel based reports into automated PDF reports to improve efficiency and reduce manual involvement
• Scorecard Development (Cart and Logistic Regression Models) for a leading US bank - SAS
• Developed custom models to predict risk at loan origination for small business lending portfolios through a score card based implementation
• OCC CCAR reporting for a leading US bank – SAS
• Prepared exhaustive reports for OCC CCAR reporting
• Prepared maker and checker mechanism for cross validating the data

May 2010 – Jul 2010 · 2 mos

• Worked on Multimedia Information Retrieval...